JP5149833B2 - Echo canceling device, echo canceling method, echo canceling program - Google Patents

Echo canceling device, echo canceling method, echo canceling program Download PDF

Info

Publication number
JP5149833B2
JP5149833B2 JP2009040644A JP2009040644A JP5149833B2 JP 5149833 B2 JP5149833 B2 JP 5149833B2 JP 2009040644 A JP2009040644 A JP 2009040644A JP 2009040644 A JP2009040644 A JP 2009040644A JP 5149833 B2 JP5149833 B2 JP 5149833B2
Authority
JP
Japan
Prior art keywords
signal
echo
adaptive filter
volume
unit
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired - Fee Related
Application number
JP2009040644A
Other languages
Japanese (ja)
Other versions
JP2010199809A (en
Inventor
翔一郎 齊藤
朗 中川
陽一 羽田
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nippon Telegraph and Telephone Corp
Original Assignee
Nippon Telegraph and Telephone Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nippon Telegraph and Telephone Corp filed Critical Nippon Telegraph and Telephone Corp
Priority to JP2009040644A priority Critical patent/JP5149833B2/en
Publication of JP2010199809A publication Critical patent/JP2010199809A/en
Application granted granted Critical
Publication of JP5149833B2 publication Critical patent/JP5149833B2/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Landscapes

  • Cable Transmission Systems, Equalization Of Radio And Reduction Of Echo (AREA)
  • Circuit For Audible Band Transducer (AREA)

Description

本発明は、収音信号から受話信号の反響を取り除くことにより送話信号を生成する反響消去装置、反響消去方法、反響消去プログラムに関する。   The present invention relates to an echo canceling apparatus, an echo canceling method, and an echo canceling program for generating a transmission signal by removing the echo of a received signal from a collected sound signal.

反響消去装置(エコーキャンセラ)は、スピーカから出て部屋のエコー経路(音響特性など)を経てマイクに入った信号を推定し、マイクで収音した信号(収音信号)から差し引いて送話信号とするものである。従来の技術とした、非特許文献1の「線形適応フィルタ」を用いた反響消去装置と、非特許文献2の「Volterraフィルタ」を用いた非線形反響消去装置がある。   The echo canceler (echo canceller) estimates the signal that enters the microphone through the room echo path (acoustic characteristics, etc.) from the speaker, and subtracts it from the signal collected by the microphone (sound collected signal). It is what. There are an echo canceling apparatus using a “linear adaptive filter” in Non-Patent Document 1 and a nonlinear echo canceling apparatus using a “Volterra filter” in Non-Patent Document 2, which are conventional techniques.

非特許文献1の反響消去装置の構成例を図1に示す。通信相手の装置から送られてきた受話信号x(n)は、スピーカ970から出力され、部屋などのエコー経路990を経てマイク980で収音される。反響消去装置900は、フィルタ手段911と更新量計算手段912を有する適応フィルタ部910と、減算部920とを備える。反響消去装置900は、受話信号x(n)とマイクで収音した収音信号y(n)とを入力とし、通信相手の装置に送る送話信号e(n)を出力する。nはデジタル信号の時刻を示す整数である。適応フィルタ部910のフィルタ手段911のタップ長をL(Lは1以上の整数)とすると、時刻nのときのフィルタ手段911の適応フィルタ係数h^(n)は、
h^(n)=[h(n),h(n),…,h(n)] (1)
と表現できる。ただし、Tは転置を意味する。このまた、現在を含めた過去L個分の受話信号x(n)のベクトル表現を
x^(n)=[x(n),x(n−1),…,x(n−L+1)] (2)
とおく。フィルタ手段911は、受話信号ベクトルx^(n)と適応フィルタ係数h^(n)から、擬似エコー信号z(n)を、
z(n)=h^(n)x^(n) (3)
のように求め出力する。減算部920は、収音信号y(n)から擬似エコー信号z(n)を減算し、
e(n)=y(n)−z(n) (4)
のように送話信号e(n)を求める。また、適応フィルタ部910が、例えばNLMSアルゴリズムによる適応フィルタの場合、更新量計算手段912は、更新量w^(n)を
A configuration example of the echo canceling apparatus of Non-Patent Document 1 is shown in FIG. The reception signal x (n) transmitted from the communication partner apparatus is output from the speaker 970 and collected by the microphone 980 via the echo path 990 such as a room. The echo cancellation apparatus 900 includes an adaptive filter unit 910 having a filter unit 911 and an update amount calculation unit 912, and a subtraction unit 920. The echo canceling apparatus 900 receives the received signal x (n) and the collected sound signal y (n) collected by the microphone, and outputs a transmitted signal e (n) to be sent to the communication partner apparatus. n is an integer indicating the time of the digital signal. When the tap length of the filter unit 911 of the adaptive filter unit 910 is L (L is an integer equal to or greater than 1), the adaptive filter coefficient h ^ (n) of the filter unit 911 at time n is
h ^ (n) = [h 1 (n), h 2 (n),..., h L (n)] T (1)
Can be expressed. However, T means transposition. Further, the vector representation of the past L received signals x (n) including the present is expressed as follows: x ^ (n) = [x (n), x (n−1),..., X (n−L + 1)] T (2)
far. The filter means 911 receives the pseudo echo signal z (n) from the received signal vector x ^ (n) and the adaptive filter coefficient h ^ (n),
z (n) = h ^ T (n) x ^ (n) (3)
Output as follows. The subtraction unit 920 subtracts the pseudo echo signal z (n) from the collected sound signal y (n),
e (n) = y (n) -z (n) (4)
The transmission signal e (n) is obtained as follows. Further, when the adaptive filter unit 910 is an adaptive filter based on the NLMS algorithm, for example, the update amount calculation unit 912 calculates the update amount w ^ (n).

Figure 0005149833
Figure 0005149833

のように求め、フィルタ手段911の次の時刻n+1の適応フィルタ係数h^(n+1)を、
h^(n+1)=h^(n)+w^(n) (6)
のように毎時刻更新する。ただし、‖・‖はベクトルのノルム、μは更新量を調整する係数(ステップサイズ、0<μ<2)、δは式(5)の分母が0にならないための微小な正の定数である。ただし、適応フィルタ部910では線形な反響しか推定しないため、送話信号には非線形な反響が残ってしまい、通話品質が劣化する可能性がある。
The adaptive filter coefficient h ^ (n + 1) at the next time n + 1 of the filter means 911 is obtained as follows:
h ^ (n + 1) = h ^ (n) + w ^ (n) (6)
Update every hour as follows. Where ‖ and ‖ are vector norms, μ is a coefficient for adjusting the amount of update (step size, 0 <μ <2), and δ is a small positive constant that prevents the denominator of Equation (5) from becoming zero. . However, since the adaptive filter unit 910 estimates only a linear echo, a non-linear echo remains in the transmitted signal, which may degrade the speech quality.

非特許文献2は、線形の適応フィルタ以外にVolterra級数の高次項に対応する適応フィルタを用意し、その入力に受話信号に処理を加えたものを用いる点が異なる。例えば、時刻nの2次のVolterraフィルタとその入力は、受話信号x(n)の過去Lサンプル分を考慮すると、
h^(n)=[h0,0(n),h0,1(n),…,h0,L−1(n),
1,1(n),…,hL−1,L−1(n)] (7)
x^(n)=[x(n),x(n)x(n−1),…,x(n)x(n−L+1),
(n−1),…,x(n−L+1)] (8)
となる。
Non-Patent Document 2 differs in that an adaptive filter corresponding to a higher-order term of the Volterra series is prepared in addition to the linear adaptive filter, and the received signal is processed at the input. For example, the second-order Volterra filter at time n and its input take into account the past L samples of the received signal x (n),
h 2 (n) = [h 0,0 (n), h 0,1 (n),..., h 0, L-1 (n),
h 1,1 (n),..., h L-1, L-1 (n)] (7)
x ^ 2 (n) = [ x 2 (n), x (n) x (n-1), ..., x (n) x (n-L + 1),
x 2 (n−1),..., x 2 (n−L + 1)] (8)
It becomes.

このフィルタは非線形時不変な入力に対応可能だが、高次項を考慮すると次数のべき乗でフィルタ長が長くなり、演算量が多くなる。また、実測では定常音には高い性能を発揮するものの、音量の変動が大きい音声では推定が不安定になり、効果が得られないことがある。   This filter can deal with non-linear time-invariant inputs, but considering higher order terms, the filter length increases with the power of the order and the amount of computation increases. In actual measurement, although high performance is exhibited for stationary sounds, estimation may become unstable and may not be effective for speech with large fluctuations in volume.

S.Haykin, “Adaptive Filter Theory,” Prentice-Hall International Inc., pp.432〜436, 1996.S. Haykin, “Adaptive Filter Theory,” Prentice-Hall International Inc., pp.432–436, 1996. A. Stenger, L. Trautmann, and R. Rabenstein, “Nonlinear acoustic echo cancellation with 2ndorder adaptive volterra filters,” Int. Conf. on Acoustics, Speech and Signal Processing, vol.2, pp877〜880, 1999.A. Stenger, L. Trautmann, and R. Rabenstein, “Nonlinear acoustic echo cancellation with 2ndorder adaptive volterra filters,” Int. Conf. On Acoustics, Speech and Signal Processing, vol.2, pp877-880, 1999.

例えばハンズフリー通話装置ではスピーカの径が小さいなどの理由から、スピーカとマイクの間のエコー経路の音響特性が非線形となる場合がある。このような場合、非特許文献1の反響消去装置ではエコー経路の線形性を仮定しているので、十分な反響消去ができないため、通話品質が劣化してしまう。また、非特許文献2の反響消去装置では、上述のように演算量が多くなることや音量の変動が大きい音声では不安定になるなどの問題がある。   For example, in a hands-free call device, the acoustic characteristics of the echo path between the speaker and the microphone may be nonlinear because the speaker diameter is small. In such a case, since the echo canceller of Non-Patent Document 1 assumes the linearity of the echo path, the echo quality cannot be satisfactorily cancelled, so that the call quality deteriorates. In addition, the echo canceller of Non-Patent Document 2 has problems such as an increase in the amount of computation as described above and instability in a voice with a large volume fluctuation.

本発明は、このような問題に鑑みてなされたものであり、通信装置の構造などの様々な理由からエコー経路の音響特性が非線形となってしまう場合でも、十分に反響を消去できる反響消去装置を提供することを目的とする。   The present invention has been made in view of such a problem, and an echo canceling apparatus capable of sufficiently canceling the echo even when the acoustic characteristics of the echo path become nonlinear due to various reasons such as the structure of the communication apparatus. The purpose is to provide.

本発明の反響消去装置は、音量判定部、M個の適応フィルタ部(ただし、Mは2以上の整数)、受話信号切替部、送話信号切替部、減算部を備える。音量判定部は、受話信号の音量があらかじめ定めたM個の音量範囲の中のどの音量範囲に属すかを示す音量情報を出力する。M個の適応フィルタ部はM個の音量情報のそれぞれに対応し、受話信号をフィルタリングすることで擬似エコー信号を生成するフィルタ手段と、受話信号と送話信号からフィルタ手段を更新する更新量計算手段とをそれぞれ有する。受話信号切替部は、受話信号が入力される適応フィルタ部を、音量判定部が出力した音量情報と対応する適応フィルタ部に切り替える。送話信号切替部は、送話信号が入力される適応フィルタ部を、音量判定部が出力した音量情報と対応する適応フィルタ部に切り替える。減算部は、収音信号から擬似エコー信号を減算する。 Anti sound erasing apparatus of the present invention comprises, volume determination section, M-number of the adaptive filter section (where, M is an integer of 2 or more), the reception signal switching unit, transmission signal switching section, a subtraction section. Volume determination section, the volume of the received signal and outputs the shown to volume information what belongs to the volume range in the M volume a predetermined range. The M adaptive filter units correspond to each of the M volume information, filter means for generating a pseudo echo signal by filtering the received signal, and update amount calculation for updating the filter means from the received signal and the transmitted signal Each means. Reception signal switching unit switches the adaptive filter unit received talk signal is input, the adaptive filter corresponding to the sound information volume determination unit is output. Transmission signal switching section switches the adaptive filter unit feed talk signal is input, the adaptive filter corresponding to the sound information volume determination unit is output. The subtracting unit subtracts the pseudo echo signal from the collected sound signal.

本発明の反響消去装置は、複数個の適応フィルタ部を備え、各適応フィルタ部が主にフィルタリングする音量がそれぞれ異なる。なお、「主にフィルタリングする音量」とは、音量情報に従って受話信号切替部と送話信号切替部とが選択した適応フィルタがフィルタリングする音量範囲(音量情報が示す音量範囲)である Anti sound erasing apparatus of the present invention comprises a plurality of adaptive filter section, the volume each adaptive filter is mainly filtering Ru different respectively. Note that "mainly Filtering volume" is a volume range adaptive filter and the received signal switching unit and the transmission signal switching section is selected according to the sound volume information is filtered (the volume range indicated volume information).

本発明の反響消去装置によれば、複数個の適応フィルタ部を備え、各適応フィルタ部が主にフィルタリングする音量がそれぞれ異なるので、受話信号の音量に依存して変動するインパルス応答を精度よく推定できる。したがって、エコー経路の音響特性が非線形な場合でも、本発明の反響消去装置は十分に反響を消去できる。また、各適応フィルタ部は線形なフィルタリングを行うので、演算量が多くなることを避けることができる。   According to the echo canceling apparatus of the present invention, a plurality of adaptive filter sections are provided, and the volume of each adaptive filter section that is mainly filtered is different, so that an impulse response that varies depending on the volume of the received signal can be accurately estimated. it can. Therefore, even when the acoustic characteristic of the echo path is nonlinear, the echo canceling apparatus of the present invention can sufficiently cancel the echo. Moreover, since each adaptive filter part performs linear filtering, it can avoid that the amount of calculations increases.

従来の反響消去装置の機能構成例を示す図。The figure which shows the function structural example of the conventional echo cancellation apparatus. 実施例1の反響消去装置の機能構成例を示す図。FIG. 3 is a diagram illustrating a functional configuration example of the echo canceling apparatus according to the first embodiment. 実施例1から3の適応フィルタ部の機能構成例を示す図。FIG. 4 is a diagram illustrating a functional configuration example of an adaptive filter unit according to the first to third embodiments. 実施例1の反響消去装置の処理フロー例を示す図。FIG. 3 is a diagram illustrating a processing flow example of the echo canceling apparatus according to the first embodiment. 実施例2の反響消去装置の機能構成例を示す図。FIG. 5 is a diagram illustrating a functional configuration example of an echo canceling apparatus according to a second embodiment. 実施例2の反響消去装置の処理フロー例を示す図。FIG. 10 is a diagram illustrating a processing flow example of the echo canceling apparatus according to the second embodiment. 実施例3の反響消去装置の機能構成例を示す図。FIG. 9 is a diagram illustrating a functional configuration example of an echo canceling apparatus according to a third embodiment. 実施例3の反響消去装置の処理フロー例を示す図。FIG. 10 is a diagram illustrating a processing flow example of the echo canceling apparatus according to the third embodiment. 実施例4の反響消去装置の機能構成例を示す図。FIG. 10 is a diagram illustrating a functional configuration example of an echo canceling apparatus according to a fourth embodiment. 実施例4から6の適応フィルタ部の機能構成例を示す図。The figure which shows the function structural example of the adaptive filter part of Examples 4-6. 実施例4の反響消去装置の処理フロー例を示す図。FIG. 10 is a diagram illustrating a processing flow example of the echo canceling apparatus according to the fourth embodiment. 実施例5の反響消去装置の機能構成例を示す図。FIG. 10 is a diagram illustrating a functional configuration example of an echo canceling apparatus according to a fifth embodiment. 実施例5の反響消去装置の処理フロー例を示す図。FIG. 10 is a diagram illustrating a processing flow example of an echo canceling apparatus according to a fifth embodiment. 実施例6の反響消去装置の機能構成例を示す図。FIG. 10 is a diagram illustrating a functional configuration example of an echo canceling apparatus according to a sixth embodiment. 実施例6の反響消去装置の処理フロー例を示す図。FIG. 10 is a diagram illustrating a processing flow example of an echo canceling apparatus according to a sixth embodiment. コンピュータの機能構成例を示す図。The figure which shows the function structural example of a computer. M系列を用いて測定した5つのインパルス応答を示す図。The figure which shows five impulse responses measured using M series. 図17の点線で囲んだ部分を拡大した図。The figure which expanded the part enclosed with the dotted line of FIG. インパルス応答1とその他のインパルス応答の振幅波形の差分を示す図。The figure which shows the difference of the amplitude waveform of the impulse response 1 and other impulse responses. 従来の反響消去装置と本発明の反響消去装置の違いを確認した実験の結果を示す図。The figure which shows the result of the experiment which confirmed the difference between the conventional echo cancellation apparatus and the echo cancellation apparatus of this invention.

以下、本発明の実施の形態について、詳細に説明する。なお、同じ機能を有する構成部には同じ番号を付し、重複説明を省略する。以降の説明では、nはデジタル信号の時刻を示す整数、Mは2以上の整数、mは1〜Mの整数、Lは2以上の整数である。x(n)は時刻nの受話信号、y(n)は時刻nの収音信号、e(n)は時刻nの送話信号である。x^(n)は、過去L個分の受話信号x(n)のベクトル表現であり、
x^(n)=[x(n),x(n−1),…,x(n−L+1)] (7)
のように表現される。ただし、Tは転置を意味する。
Hereinafter, embodiments of the present invention will be described in detail. In addition, the same number is attached | subjected to the structure part which has the same function, and duplication description is abbreviate | omitted. In the following description, n is an integer indicating the time of the digital signal, M is an integer of 2 or more, m is an integer of 1 to M, and L is an integer of 2 or more. x (n) is a reception signal at time n, y (n) is a sound pickup signal at time n, and e (n) is a transmission signal at time n. x ^ (n) is a vector representation of the past L received signals x (n),
x ^ (n) = [x (n), x (n−1),..., x (n−L + 1)] T (7)
It is expressed as However, T means transposition.

図2に実施例1の反響消去装置の機能構成例を、図3に適応フィルタ部の機能構成例を示す。また、図4に実施例1の反響消去装置の処理フロー例を示す。反響消去装置100は、音量判定部130、タップ長がLのM個の適応フィルタ部110−1,…,110−M、受話信号切替部140、送話信号切替部150、減算部920を備える。図3(A)に示した適応フィルタ部110−mは、フィルタ手段111−mと更新量計算手段112−mとをそれぞれ有する。なお、フィルタ手段111−mの適応フィルタ係数h^(n)は、
^(n)=[h1,m(n),h2,m(n),…,hL,m(n)] (8)
と表現できる。
FIG. 2 shows a functional configuration example of the echo canceling apparatus according to the first embodiment, and FIG. 3 shows a functional configuration example of the adaptive filter unit. FIG. 4 shows a processing flow example of the echo canceling apparatus of the first embodiment. The echo cancellation apparatus 100 includes a volume determination unit 130, M adaptive filter units 110-1,..., 110-M having a tap length L, a received signal switching unit 140, a transmitted signal switching unit 150, and a subtraction unit 920. . The adaptive filter unit 110-m illustrated in FIG. 3A includes a filter unit 111-m and an update amount calculation unit 112-m. The adaptive filter coefficient h m ^ (n) of the filter means 111-m is
h m ^ (n) = [h 1, m (n), h 2, m (n),..., h L, m (n)] T (8)
Can be expressed.

音量判定部130は、受話信号x(n)の音量があらかじめ定めたM個の音量範囲の中のどの音量範囲に属すかを判定し、該当する音量範囲を示す音量情報mを出力する(S130)。具体的には、以下のように行えばよいがこの方法に限定する必要はない。あらかじめM−1個の閾値v,…,vM−1(ただし、v<v<…<vM−1)を定めておく。そして、受話信号x(n)の音量として受話信号ベクトルx^(n)のノルム‖x^(n)‖を求め、 The volume determination unit 130 determines which volume range of the M volume ranges the volume of the received signal x (n) belongs to, and outputs volume information m indicating the corresponding volume range (S130). ). Specifically, it may be performed as follows, but it is not necessary to limit to this method. M−1 threshold values v 1 ,..., V M−1 (where v 1 <v 2 <... <V M−1 ) are determined in advance. Then, the norm ‖x ^ (n) ‖ of the received signal vector x ^ (n) is obtained as the volume of the received signal x (n),

Figure 0005149833
Figure 0005149833

のように音量情報mを求める。このように音量情報mを求めた場合、mが大きくなるほど大きい音量に対応することになる。なお、受話信号x(n)の音量は、時刻n−L+1〜nの間の受話信号x(n)の最大値としてもよいし、要素の数をL(適応フィルタ部のタップ長)とは異なる数とした受話信号ベクトルのノルムとしてもよい。 The volume information m is obtained as follows. When the volume information m is obtained in this way, the larger the m, the higher the volume. The volume of the received signal x (n) may be the maximum value of the received signal x (n) between times n−L + 1 and n, and the number of elements is L (tap length of the adaptive filter unit). The norm of the received signal vector having a different number may be used.

受話信号切替部140は、音量情報mに従って受話信号x(n)を、いずれか1つの適応フィルタ部110−mに入力する(S140)。送話信号切替部150は、音量情報mに従って送話信号e(n)を、受話信号切替部140が受話信号x(n)を入力した適応フィルタ部110−mに入力する(S150)。   The received signal switching unit 140 inputs the received signal x (n) according to the volume information m to any one of the adaptive filter units 110-m (S140). The transmission signal switching unit 150 inputs the transmission signal e (n) according to the volume information m to the adaptive filter unit 110-m to which the reception signal switching unit 140 has input the reception signal x (n) (S150).

適応フィルタ部110−mのフィルタ手段111−mは、受話信号x(n)をフィルタリングすることで擬似エコー信号z(n)を生成する(S111)。具体的には、適応フィルタ係数h^(n)と受話信号ベクトルx^(n)を用いて、 The filter unit 111-m of the adaptive filter unit 110-m generates the pseudo echo signal z (n) by filtering the received signal x (n) (S111). Specifically, using the adaptive filter coefficient h m ^ (n) and the received signal vector x ^ (n),

Figure 0005149833
Figure 0005149833

のように求めればよい。減算部920は、収音信号y(n)から擬似エコー信号z(n)を減算し、
e(n)=y(n)−z(n) (11)
のように送話信号e(n)を求める(S920)。
You can ask as follows. The subtraction unit 920 subtracts the pseudo echo signal z (n) from the collected sound signal y (n),
e (n) = y (n) -z (n) (11)
The transmission signal e (n) is obtained as follows (S920).

更新量計算手段112−mは、受話信号x(n)と送話信号e(n)からフィルタ手段111−mの適応フィルタ係数h^(n)を更新して次の時刻n+1の適応フィルタ係数h^(n+1)を求める(S112)。具体的には、更新量計算手段112−mは、更新量w^(n)を The update amount calculating means 112-m updates the adaptive filter coefficient h m ^ (n) of the filter means 111-m from the received signal x (n) and the transmitted signal e (n), and the adaptive filter at the next time n + 1. A coefficient h m ^ (n + 1) is obtained (S112). Specifically, the update amount calculation unit 112-m calculates the update amount w ^ (n).

Figure 0005149833
Figure 0005149833

のように求め、フィルタ手段111−mの次の時刻n+1の適応フィルタ係数h^(n+1)を、
^(n+1)=h^(n)+w^(n) (13)
のように毎時刻更新する。ただし、‖・‖はベクトルのノルム、μは更新量を調整する係数(ステップサイズ、0<μ<2)、δは式(13)の分母が0にならないための微小な正の定数である。
The adaptive filter coefficient h m ^ (n + 1) at the next time n + 1 of the filter means 111-m is obtained as follows:
h m ^ (n + 1) = h m ^ (n) + w ^ (n) (13)
Update every hour as follows. Where ‖ and ‖ are vector norms, μ is a coefficient for adjusting the update amount (step size, 0 <μ <2), and δ is a small positive constant that prevents the denominator of Equation (13) from becoming zero. .

反響消去装置100は、このような処理を時刻ごとに繰り返す。反響消去装置100では、各時刻で受話信号x(n)の音量に従って適応フィルタ部を選択するので、適応フィルタ係数h^(n)が更新される適応フィルタ部110−mも1つだけである。つまり、1つの適応フィルタ部110−mに着目すると、特定の音量範囲のときだけ受話信号x(n)が入力され、適応フィルタ係数h^(n)が更新される。したがって、その適応フィルタ部110−mの適応フィルタ係数h^(n)は、次第にその適応フィルタ部110−mがフィルタリングする音量範囲に適合したものに更新されていく。そして、特定の音量範囲に適合した適応フィルタ部110−mだけを受話信号の音量に従って選定し、フィルタリングする。したがって、エコー経路990の音響特性が音量に依存する非線形性を有していても、的確に擬似エコー信号を推定できるので、十分に反響を消去できる。また、各時刻では、線形なフィルタリングを行う1つの適応フィルタ部がフィルタリングするだけなので、演算量が多くなることを避けることができる。 The echo canceling apparatus 100 repeats such processing for each time. Since the echo canceling apparatus 100 selects an adaptive filter unit according to the volume of the received signal x (n) at each time, the adaptive filter unit 110-m whose adaptive filter coefficient h m ^ (n) is updated is only one. is there. That is, paying attention to one adaptive filter unit 110-m, the received signal x (n) is input only in a specific volume range, and the adaptive filter coefficient h m ^ (n) is updated. Therefore, the adaptive filter coefficient h m ^ (n) of the adaptive filter unit 110-m is gradually updated to be adapted to the volume range filtered by the adaptive filter unit 110-m. Then, only the adaptive filter unit 110-m suitable for the specific volume range is selected and filtered according to the volume of the received signal. Therefore, even if the acoustic characteristic of the echo path 990 has nonlinearity that depends on the volume, the pseudo echo signal can be accurately estimated, so that the echo can be sufficiently eliminated. In addition, at each time, since only one adaptive filter unit that performs linear filtering performs filtering, an increase in the amount of calculation can be avoided.

なお、適応フィルタ部110−mは、周波数領域での処理を行ってもよい。例えば、図3(B)に示したように、適応フィルタ部110−mは、周波数変換手段113−m、115−m、周波数逆変換手段114−m、周波数領域の処理を行うフィルタ手段116−m、周波数領域の処理を行う更新量計算手段117−mを備えてもよい。図3(B)の適応フィルタ部110−mの場合は、受話信号x(n)は、いくつかの受話信号x(n),…,x(n−F+1)ごとに周波数変換手段113−mで受話信号スペクトルX^(k)に変換される。ただし、Fは1フレームを構成する受話信号x(n)の数、kは周波数を示すインデックスである。フィルタ手段116−mはフィルタリングによって擬似エコー信号スペクトルY^(k)を求める。周波数逆変換手段114−mは擬似エコー信号スペクトルY^(k)を擬似エコー信号y(n),…,y(n−F+1)に変換する。また、周波数変換手段115−mは、いくつかの送話信号e(n),…,e(n−F+1)ごとに送話信号スペクトルE^(k)に変換する。更新量計算手段117−mは、受話信号スペクトルX^(k)と送話信号スペクトルE^(k)から更新量W^(k)を求め、フィルタ手段116−mの適応フィルタ係数を更新する。このような周波数領域での処理を行う適応フィルタを用いても、同様の効果が得られる。   Note that the adaptive filter unit 110-m may perform processing in the frequency domain. For example, as illustrated in FIG. 3B, the adaptive filter unit 110-m includes frequency conversion units 113-m and 115-m, a frequency inverse conversion unit 114-m, and a filter unit 116- that performs frequency domain processing. m, update amount calculation means 117-m for performing processing in the frequency domain may be provided. In the case of the adaptive filter section 110-m in FIG. 3B, the received signal x (n) is converted into frequency conversion means 113-m for each of several received signals x (n),..., X (n−F + 1). Is converted into the received signal spectrum X ^ (k). Here, F is the number of received signals x (n) constituting one frame, and k is an index indicating the frequency. The filter means 116-m obtains a pseudo echo signal spectrum Y ^ (k) by filtering. The frequency inverse conversion means 114-m converts the pseudo echo signal spectrum Y ^ (k) into pseudo echo signals y (n),..., Y (n−F + 1). Further, the frequency conversion means 115-m converts the transmission signal spectrum E ^ (k) for every several transmission signals e (n),..., E (n−F + 1). The update amount calculation means 117-m obtains the update amount W ^ (k) from the reception signal spectrum X ^ (k) and the transmission signal spectrum E ^ (k), and updates the adaptive filter coefficient of the filter means 116-m. . The same effect can be obtained even when an adaptive filter that performs processing in such a frequency domain is used.

[変形例]
実施例1では、式(12)に示すように、すべての適応フィルタ部110−1,…,110−Mで、更新量w^(n)を求める式は同じであった。本変形例の適応フィルタ部110’−1,…,110’−Mでは、更新量を調整する係数μ(ステップサイズ)を適応フィルタ部110’−mごとに設定する。この点のみが実施例1と異なる。具体的には、μ,…,μをあらかじめ用意しておき、更新量計算手段112’−mが更新量w^(n)を
[Modification]
In the first embodiment, as shown in the equation (12), the equation for obtaining the update amount ^ (n) is the same in all the adaptive filter units 110-1,..., 110-M. In the adaptive filter units 110′-1,..., 110′-M of this modification, a coefficient μ (step size) for adjusting the update amount is set for each adaptive filter unit 110′-m. Only this point is different from the first embodiment. Specifically, μ 1 ,..., Μ M are prepared in advance, and the update amount calculation means 112′-m determines the update amount w ^ (n).

Figure 0005149833
Figure 0005149833

のように求め、フィルタ手段111−mの次の時刻n+1の適応フィルタ係数h^(n+1)を、
^(n+1)=h^(n)+w^(n) (15)
のように毎時刻更新する。ここで、フィルタリングする受話信号の音量が大きい適応フィルタ部110’−mほど、更新量計算手段112’−mが求める更新量を調整する係数μを、大きい値に設定しておく。つまり、式(9)のように音量情報mを求める場合は、μ<μ<…<μのようにμ,…,μを設定することになる。
The adaptive filter coefficient h m ^ (n + 1) at the next time n + 1 of the filter means 111-m is obtained as follows:
h m ^ (n + 1) = h m ^ (n) + w ^ (n) (15)
Update every hour as follows. Here, as the adaptive filter 110 '-m loudest to filter the received signal, the coefficient mu m to adjust the update amount of the update amount calculating means 112'-m is determined, it is set to a large value. That is, the case of obtaining the volume information m as in equation (9), mu 1 as μ 1 <μ 2 <... < μ M, ..., will set the mu M.

このように更新量を調整する係数を設定することで、実施例1と同様の効果が得られる他に、次のような効果も得られる。受話信号x(n)の音量が小さくSN比が悪い場合には、更新量w^(n)が小さくなるので反響消去量が低下しないようにできる。また、受話信号x(n)の音量が大きくSN比がよい場合には、更新量w^(n)が大きくなるので収束速度を速くできる。   By setting the coefficient for adjusting the update amount in this way, the same effect as in the first embodiment can be obtained, and the following effect can also be obtained. When the volume of the received signal x (n) is small and the S / N ratio is bad, the update amount w ^ (n) is small, so that the echo cancellation amount can be prevented from decreasing. Further, when the volume of the received signal x (n) is large and the SN ratio is good, the update amount w ^ (n) becomes large, so that the convergence speed can be increased.

図5に実施例2の反響消去装置の機能構成例を、図6に実施例2の反響消去装置の処理フロー例を示す。実施例2の反響消去装置200は、固定フィルタ部210と加算部260が追加された点が反響消去装置100と異なる。固定フィルタ部210は、受話信号x(n)をフィルタリングし、固定擬似エコー信号z(n)を出力する(S210)。例えば、固定フィルタ部210のタップ長をL’(ただし、L’は2以上の整数)、固定フィルタ係数h^を
^=[h1,f,h2,f,…,hL’,f] (16)
とすると、固定フィルタ部210から出力される固定擬似エコー信号z(n)は、
FIG. 5 shows a functional configuration example of the echo canceling apparatus of the second embodiment, and FIG. 6 shows a processing flow example of the echo canceling apparatus of the second embodiment. The echo canceling apparatus 200 according to the second embodiment is different from the echo canceling apparatus 100 in that a fixed filter unit 210 and an adding unit 260 are added. The fixed filter unit 210 filters the received signal x (n) and outputs a fixed pseudo echo signal z f (n) (S210). For example, the tap length of the fixed filter unit 210 is L ′ (where L ′ is an integer equal to or greater than 2), and the fixed filter coefficient h f ^ is h f ^ = [h 1, f , h 2, f ,..., H L ', F ] T (16)
Then, the fixed pseudo echo signal z f (n) output from the fixed filter unit 210 is

Figure 0005149833
Figure 0005149833

となる。なお、固定フィルタ係数h^は、無響室であらかじめ測定しておいたある音量でのインパルス応答などを用いればよい。 It becomes. Note that the fixed filter coefficient h f ^ may be an impulse response at a certain volume measured in advance in an anechoic chamber.

加算部260は、適応フィルタ部110−mが出力した擬似エコー信号と固定擬似エコー信号z(n)とを加算した信号を、新たな擬似エコー信号z(n)とする(S260)。したがって、減算部920に入力される擬似エコー信号z(n)は、 The adding unit 260 sets a signal obtained by adding the pseudo echo signal output from the adaptive filter unit 110-m and the fixed pseudo echo signal z f (n) as a new pseudo echo signal z (n) (S260). Therefore, the pseudo echo signal z (n) input to the subtraction unit 920 is

Figure 0005149833
Figure 0005149833

となる。その他の処理フローは実施例1と同じである。また、実施例1変形例のように、更新量を調整する係数がそれぞれ異なる適応フィルタ部110’−1,…,110’−Mを用いてもよい。 It becomes. Other processing flows are the same as those in the first embodiment. Further, as in the first embodiment, adaptive filter units 110'-1, ..., 110'-M having different coefficients for adjusting the update amount may be used.

反響消去装置200は固定フィルタ部210を備えているので、M個の適応フィルタ部110−1,…,110−Mは、固定フィルタ係数を求めたときの条件との差によって生じる反響の変化分を推定することになる。例えば、固定フィルタ係数が、無響室であらかじめ測定しておいたある音量でのインパルス応答の場合、実際の部屋での反響の違いと受話信号の音量の違いによるインパルス応答の変化のみを推定することになる。このインパルス応答の変化が、固定フィルタ部210のタップ長L’よりも短いタップ数で表現できる場合、適応フィルタ部110−mのタップ数LをL’よりも小さくできる。したがって、反響消去装置200では、実施例1や実施例1変形例と同様の効果が得られる他に、適応フィルタ係数h^(n)の要素の数を少なくでき、演算量を減らすことも期待できる。 Since the echo canceling apparatus 200 includes the fixed filter unit 210, the M adaptive filter units 110-1,..., 110-M have the amount of change in the echo caused by the difference from the condition when the fixed filter coefficient is obtained. Will be estimated. For example, if the fixed filter coefficient is an impulse response at a certain volume measured in an anechoic room in advance, only the change in the impulse response due to the difference in the echo in the actual room and the volume of the received signal is estimated. It will be. When the change in the impulse response can be expressed by a tap number shorter than the tap length L ′ of the fixed filter unit 210, the tap number L of the adaptive filter unit 110-m can be made smaller than L ′. Therefore, in the echo canceling apparatus 200, in addition to the same effects as those of the first embodiment and the first embodiment, the number of elements of the adaptive filter coefficient h m ^ (n) can be reduced, and the calculation amount can be reduced. I can expect.

図7に実施例3の反響消去装置の機能構成例を、図8に実施例3の反響消去装置の処理フロー例を示す。実施例3の反響消去装置300は、遅延部370、適応フィルタ部310、加算部360が追加された点が反響消去装置100と異なる。また、適応フィルタ部310の機能構成例は図3に示す。   FIG. 7 shows a functional configuration example of the echo canceling apparatus of the third embodiment, and FIG. 8 shows an example of a processing flow of the echo canceling apparatus of the third embodiment. The echo canceling apparatus 300 according to the third embodiment is different from the echo canceling apparatus 100 in that a delay unit 370, an adaptive filter unit 310, and an adding unit 360 are added. An example of the functional configuration of the adaptive filter unit 310 is shown in FIG.

遅延部370は、受話信号x(n)をあらかじめ定めた時間Dだけ遅延させて、遅延信号x(n−D)を出力する(S370)。適応フィルタ部310は、フィルタ手段311(316)と更新量計算手段312(317)とを有する。適応フィルタ部310のフィルタ手段311は、遅延信号x(n−D)をフィルタリングすることで遅延擬似エコー信号z(n)を生成する(S311)。例えば、適応フィルタ部310のタップ長をL”(ただし、L”は2以上の整数)、適応フィルタ係数h^を
^(n)=[h1,0(n),h2,0(n),…,hL”,0(n)] (19)
とすると、適応フィルタ部310から出力される遅延擬似エコー信号z(n)は、
The delay unit 370 delays the received signal x (n) by a predetermined time D and outputs a delayed signal x (n−D) (S370). The adaptive filter unit 310 includes filter means 311 (316) and update amount calculation means 312 (317). The filter unit 311 of the adaptive filter unit 310 generates the delayed pseudo echo signal z d (n) by filtering the delayed signal x (n−D) (S311). For example, the tap length of the adaptive filter unit 310 is L ″ (where L ″ is an integer greater than or equal to 2), and the adaptive filter coefficient h 0 ^ is set to h 0 ^ (n) = [h 1,0 (n), h 2, 0 (n),..., H L ″, 0 (n)] T (19)
Then, the delayed pseudo echo signal z d (n) output from the adaptive filter unit 310 is

Figure 0005149833
Figure 0005149833

となる。
加算部360は、適応フィルタ部110−mが出力した擬似エコー信号と遅延擬似エコー信号z(n)とを加算した信号を、新たな擬似エコー信号z(n)とする(S360)。したがって、減算部920に入力される擬似エコー信号z(n)は、
It becomes.
The adding unit 360 sets a signal obtained by adding the pseudo echo signal output from the adaptive filter unit 110-m and the delayed pseudo echo signal z d (n) as a new pseudo echo signal z (n) (S360). Therefore, the pseudo echo signal z (n) input to the subtraction unit 920 is

Figure 0005149833
Figure 0005149833

となる。適応フィルタ部310の更新量計算手段312は、受話信号x(n)と送話信号e(n)からフィルタ手段311を更新する(S312)。具体的には、更新量計算手段312は、更新量w^(n)を It becomes. The update amount calculation unit 312 of the adaptive filter unit 310 updates the filter unit 311 from the received signal x (n) and the transmitted signal e (n) (S312). Specifically, the update amount calculation means 312 calculates the update amount w ^ (n).

Figure 0005149833
Figure 0005149833

のように求め、フィルタ手段111−mの次の時刻n+1の適応フィルタ係数h^(n+1)を、
^(n+1)=h^(n)+w^(n) (23)
のように毎時刻更新する。その他の処理フローは実施例1と同じである。また、実施例1変形例のように、更新量を調整する係数がそれぞれ異なる適応フィルタ部110’−1,…,110’−Mを用いてもよい。
And the adaptive filter coefficient h 0 ^ (n + 1) at the next time n + 1 of the filter means 111-m is
h 0 ^ (n + 1) = h 0 ^ (n) + w ^ (n) (23)
Update every hour as follows. Other processing flows are the same as those in the first embodiment. Further, as in the first embodiment, adaptive filter units 110′-1,..., 110′-M having different coefficients for adjusting the update amount may be used.

反響消去装置300では、L+L”が実質的なタップ長となる。系の非線形性の影響がフィルタのはじめのL個までにしか含まれていない場合、非線形性に対応するためにM個備えられている適応フィルタ部110−mのタップ長Lを短くできる。そして、音量に依存しない線形な範囲を適応フィルタ部310で対応すればよい。このようにLとL”とを定めれば、反響消去の性能は反響消去装置100と同等にできる。さらに、適応フィルタ係数を記録するメモリは、M×L+L”必要であるが、Lを短くできるので全体としてはメモリを節約できる。例えば、ΔLだけ短くできた場合、反響消去装置100の場合よりも、M×ΔL−L”だけメモリを節約できる。   In the echo canceller 300, L + L ″ is a substantial tap length. When the influence of the nonlinearity of the system is included only in the first L of the filter, M is provided to cope with the nonlinearity. The tap length L of the adaptive filter unit 110-m can be shortened, and a linear range that does not depend on the volume may be dealt with by the adaptive filter unit 310. If L and L ″ are determined in this way, the echo will be reflected. The erasing performance can be equivalent to the echo canceling apparatus 100. Further, the memory for recording the adaptive filter coefficient requires M × L + L ″, but L can be shortened, so that the memory can be saved as a whole. , M × ΔL−L ″ can save memory.

図9に実施例4の反響消去装置の機能構成例を、図10に適応フィルタ部の機能構成例を示す。また、図11に実施例4の反響消去装置の処理フロー例を示す。反響消去装置400は、係数計算部430、タップ長がLのM個の適応フィルタ部410−1,…,410−M、重み付け加算部461、減算部920を備える。図10(A)に示した適応フィルタ部410−mは、フィルタ手段111−mと更新量計算手段412−mとをそれぞれ有する。   FIG. 9 illustrates a functional configuration example of the echo canceling apparatus according to the fourth embodiment, and FIG. 10 illustrates a functional configuration example of the adaptive filter unit. FIG. 11 shows a processing flow example of the echo canceling apparatus of the fourth embodiment. The echo cancellation apparatus 400 includes a coefficient calculation unit 430, M adaptive filter units 410-1,..., 410-M having a tap length of L, a weighting addition unit 461, and a subtraction unit 920. The adaptive filter unit 410-m illustrated in FIG. 10A includes a filter unit 111-m and an update amount calculation unit 412-m.

係数計算部430は、受話信号x(n)の音量に基づいて、M個の重み係数β,…,βを求める(S430)。重み係数βは、適応フィルタ部410−mごとに対応している。具体的には、以下のように行えばよいがこの方法に限定する必要はない。あらかじめM個の閾値v,…,v(ただし、v<v<…<v)を定めておく。そして、受話信号x(n)の音量として受話信号ベクトルx^(n)のノルム‖x^(n)‖を求め、 The coefficient calculation unit 430 obtains M weight coefficients β 1 ,..., Β M based on the volume of the received signal x (n) (S430). The weight coefficient β m corresponds to each adaptive filter unit 410-m. Specifically, it may be performed as follows, but it is not necessary to limit to this method. M threshold values v 1 ,..., V M (where v 1 <v 2 <... <V M ) are determined in advance. Then, the norm ‖x ^ (n) ‖ of the received signal vector x ^ (n) is obtained as the volume of the received signal x (n),

Figure 0005149833
Figure 0005149833

のように重み係数β,…,βを求める。なお、受話信号x(n)の音量は、時刻n−L+1〜nの間の受話信号x(n)の最大値としてもよいし、要素の数をL(適応フィルタ部のタップ長)とは異なる数とした受話信号ベクトルのノルムとしてもよい。 The weight coefficients β 1 ,..., Β M are obtained as follows. The volume of the received signal x (n) may be the maximum value of the received signal x (n) between times n−L + 1 and n, and the number of elements is L (tap length of the adaptive filter unit). The norm of the received signal vector having a different number may be used.

各適応フィルタ部410−mのフィルタ手段111−mは、受話信号x(n)をフィルタリングすることで個別擬似エコー信号z(n)を生成する(S110−1,…,S110−M)。重み付け加算部461は、適応フィルタ部410−1,…,410−Mが生成する個別擬似エコー信号z(n),…,z(n)を、重み係数β,…,βを用いて重み付け加算して擬似エコー信号z(n)を求める(S461)。例えば、重み付け加算部461は、係数乗算部440−1,…,440−Mと加算部460で構成すればよい。係数乗算部440−mは、入力された個別擬似エコー信号z(n)と重み係数βとを乗算し、β(n)を出力する(S440−1,…,S440−M)。これらの乗算結果を、加算部460が加算することで、擬似エコー信号z(n)を The filter means 111-m of each adaptive filter unit 410-m generates the individual pseudo echo signal z m (n) by filtering the received signal x (n) (S110-1,..., S110-M). The weighting addition unit 461 uses the individual pseudo echo signals z 1 (n), ..., z M (n) generated by the adaptive filter units 410-1, ..., 410-M and the weighting coefficients β 1 , ..., β M , respectively. The pseudo echo signal z (n) is obtained by using the weighted addition (S461). For example, the weighting addition unit 461 may be configured by coefficient multiplication units 440-1,... 440-M and an addition unit 460. The coefficient multiplier 440-m multiplies the input individual pseudo echo signal z m (n) and the weight coefficient β m and outputs β m z m (n) (S440-1,..., S440-M). ). The addition unit 460 adds these multiplication results, so that the pseudo echo signal z (n) is obtained.

Figure 0005149833
Figure 0005149833

のように求める(S460)。減算部920は、収音信号y(n)から擬似エコー信号z(n)を減算し、送話信号e(n)を求める(S920)。 (S460). The subtractor 920 subtracts the pseudo echo signal z (n) from the collected sound signal y (n) to obtain a transmission signal e (n) (S920).

更新量計算手段412−mは、受話信号x(n)と送話信号e(n)と重み係数βからフィルタ手段111−mの適応フィルタ係数h^(n)を更新して次の時刻n+1の適応フィルタ係数h^(n+1)を求める(S412)。具体的には、更新量計算手段412−mは、更新量w^(n)を The update amount calculation means 412-m updates the adaptive filter coefficient h m ^ (n) of the filter means 111-m from the received signal x (n), the transmission signal e (n), and the weight coefficient β m to An adaptive filter coefficient h m ^ (n + 1) at time n + 1 is obtained (S412). Specifically, the update amount calculation means 412-m calculates the update amount w ^ (n).

Figure 0005149833
Figure 0005149833

のように求め、フィルタ手段111−mの次の時刻n+1の適応フィルタ係数h^(n+1)を、
^(n+1)=h^(n)+w^(n) (27)
のように毎時刻更新する。ただし、‖・‖はベクトルのノルム、μは更新量を調整する係数(ステップサイズ、0<μ<2)、δは式(26)の分母が0にならないための微小な正の定数である。
The adaptive filter coefficient h m ^ (n + 1) at the next time n + 1 of the filter means 111-m is obtained as follows:
h m ^ (n + 1) = h m ^ (n) + w ^ (n) (27)
Update every hour as follows. Where ‖ and ‖ are vector norms, μ is a coefficient for adjusting the update amount (step size, 0 <μ <2), and δ is a small positive constant that prevents the denominator of Equation (26) from becoming zero. .

反響消去装置400は、このような処理を時刻ごとに繰り返す。反響消去装置400では、各時刻で受話信号x(n)の音量に従って適応フィルタ部の重み係数βを計算し、重み係数βに応じた更新量で適応フィルタ係数h^(n)が更新される。つまり、1つの適応フィルタ部110−mに着目すると、重み係数βが大きくなる音量範囲のときには、適応フィルタ係数h^(n)の更新量が大きくなる。したがって、その適応フィルタ部110−mの適応フィルタ係数h^(n)は、次第にその適応フィルタ部110−mが主にフィルタリングする音量範囲に適合したものに更新されていく。そして、特定の音量範囲に適合した適応フィルタ部110−mからの出力の重み係数βが大きくなるように重み付け加算する。したがって、エコー経路990の音響特性が音量に依存する非線形性を有していても、的確に擬似エコー信号を推定できるので、十分に反響を消去できる。なお、いずれか1つの適応フィルタ部に対する重み係数を1とし、他の重み係数を0とするように重み係数を求めると、実質的に反響消去装置100と同じとなる。 The echo canceling apparatus 400 repeats such processing for each time. The echo canceling apparatus 400 calculates the weighting factor β m of the adaptive filter unit according to the volume of the received signal x (n) at each time, and the adaptive filter factor h m ^ (n) is updated with an update amount corresponding to the weighting factor β m. Updated. That is, paying attention to one adaptive filter unit 110-m, the update amount of the adaptive filter coefficient h m ^ (n) increases in the volume range where the weight coefficient β m increases. Therefore, the adaptive filter coefficient h m ^ (n) of the adaptive filter unit 110-m is gradually updated to a value suitable for the volume range that the adaptive filter unit 110-m mainly filters. The weighted addition so that the weighting coefficient beta m of the output from a specific adaptive filter unit 110-m suitable for the volume range becomes large. Therefore, even if the acoustic characteristic of the echo path 990 has nonlinearity that depends on the volume, the pseudo echo signal can be accurately estimated, so that the echo can be sufficiently eliminated. Note that if the weighting factor is determined so that the weighting factor for any one of the adaptive filter units is 1, and the other weighting factors are 0, the result is substantially the same as that of the echo canceling apparatus 100.

反響消去装置400と反響消去装置100とは、複数個の適応フィルタ部を備え、各適応フィルタ部が主にフィルタリングする音量がそれぞれ異なる点で共通する。ここで、「主にフィルタリングする音量」とは、反響消去装置100の場合には、音量情報mに従って受話信号切替部140と送話信号切替部150とが選択した適応フィルタ部110−mがフィルタリングする音量(音量情報が示す音量範囲)である。また、反響消去装置400の場合には、適応フィルタ部410−mの出力(個別擬似エコー信号)に対する重み係数βが最も大きくなる音量範囲に相当する。 The echo canceling apparatus 400 and the echo canceling apparatus 100 have a plurality of adaptive filter units, and are common in that the volume that each adaptive filter unit mainly filters differs. Here, in the case of the echo canceling apparatus 100, “mainly volume to be filtered” is filtered by the adaptive filter unit 110-m selected by the reception signal switching unit 140 and the transmission signal switching unit 150 according to the volume information m. The volume to be played (the volume range indicated by the volume information). Further, in the case of the echo canceller 400, this corresponds to a volume range in which the weighting coefficient βm for the output (individual pseudo echo signal) of the adaptive filter unit 410- m is the largest.

なお、適応フィルタ部410−mは、周波数領域での処理を行ってもよい。例えば、図10(B)に示したように、適応フィルタ部410−mは、周波数変換手段113−m、115−m、周波数逆変換手段114−m、周波数領域の処理を行うフィルタ手段116−m、周波数領域の処理を行う更新量計算手段417−mを備えてもよい。図10(B)の適応フィルタ部410−mの場合は、受話信号x(n)は、いくつかの受話信号x(n),…,x(n−F+1)ごとに周波数変換手段113−mで受話信号スペクトルX^(k)に変換される。ただし、Fは1フレームを構成する受話信号x(n)の数、kは周波数を示すインデックスである。フィルタ手段116−mはフィルタリングによって擬似エコー信号スペクトルY^(k)を求める。周波数逆変換手段114−mは擬似エコー信号スペクトルY^(k)を擬似エコー信号y(n),…,y(n−F+1)に変換する。また、周波数変換手段115−mは、いくつかの送話信号e(n),…,e(n−F+1)ごとに送話信号スペクトルE^(k)に変換する。更新量計算手段417−mは、受話信号スペクトルX^(k)と送話信号スペクトルE^(k)と重み係数βから更新量W^(k)を求め、フィルタ手段116−mの適応フィルタ係数を更新する。このような周波数領域での処理を行う適応フィルタを用いても、同様の効果が得られる。 Note that the adaptive filter unit 410-m may perform processing in the frequency domain. For example, as shown in FIG. 10B, the adaptive filter unit 410-m includes frequency conversion units 113-m and 115-m, frequency inverse conversion unit 114-m, and filter unit 116- that performs frequency domain processing. m, update amount calculation means 417-m for performing processing in the frequency domain may be provided. In the case of the adaptive filter unit 410-m in FIG. 10B, the received signal x (n) is converted into frequency conversion means 113-m for each of several received signals x (n), ..., x (n-F + 1). Is converted into the received signal spectrum X ^ (k). Here, F is the number of received signals x (n) constituting one frame, and k is an index indicating the frequency. The filter means 116-m obtains a pseudo echo signal spectrum Y ^ (k) by filtering. The frequency inverse conversion means 114-m converts the pseudo echo signal spectrum Y ^ (k) into pseudo echo signals y (n),..., Y (n−F + 1). Further, the frequency conversion means 115-m converts the transmission signal spectrum E ^ (k) for every several transmission signals e (n),..., E (n−F + 1). Update amount calculation unit 417-m is the received signal spectrum X ^ (k) an update quantity W ^ seeking (k) transmission signal spectrum E ^ (k) and the weighting coefficient beta m, the adaptive filter means 116-m Update filter coefficients. The same effect can be obtained even when an adaptive filter that performs processing in such a frequency domain is used.

[変形例1]
実施例4では、式(26)に示すように、適応フィルタ部410−1,…,410−Mの更新量w^(n)は重み係数βには依存しているが、主にフィルタリングする音量は考慮していなかった。本変形例の適応フィルタ部410’−1,…,410’−Mでは、更新量を調整する係数μ(ステップサイズ)を適応フィルタ部410’−mごとに設定する。この点のみが実施例4と異なる。具体的には、μ,…,μをあらかじめ用意しておき、更新量計算手段412’−mが更新量w^(n)を
[Modification 1]
In Example 4, as shown in equation (26), the adaptive filter 410-1, ..., the update amount of 410-M w ^ (n) but is dependent on the weighting factor beta m, primarily filtering I did not consider the volume to play. In the adaptive filter units 410′-1,..., 410′-M of this modification, a coefficient μ (step size) for adjusting the update amount is set for each adaptive filter unit 410′-m. Only this point is different from the fourth embodiment. Specifically, μ 1 ,..., Μ M are prepared in advance, and the update amount calculation means 412′-m determines the update amount w ^ (n).

Figure 0005149833
Figure 0005149833

のように求め、フィルタ手段411−mの次の時刻n+1の適応フィルタ係数h^(n+1)を、
^(n+1)=h^(n)+w^(n) (29)
のように毎時刻更新する。ここで、フィルタリングする受話信号の音量が大きい適応フィルタ部410’−mほど、更新量計算手段412’−mが求める更新量を調整する係数μを、大きい値に設定しておく。つまり、式(24)のように音量情報mを求める場合は、μ<μ<…<μのようにμ,…,μを設定することになる。
The adaptive filter coefficient h m ^ (n + 1) at the next time n + 1 of the filter means 411-m is obtained as follows:
h m ^ (n + 1) = h m ^ (n) + w ^ (n) (29)
Update every hour as follows. Here, as the adaptive filter 410 '-m loudest to filter the received signal, the coefficient mu m to adjust the update amount of the update amount calculating means 412'-m is determined, it is set to a large value. That is, the case of obtaining the volume information m as in equation (24), mu 1 as μ 1 <μ 2 <... < μ M, ..., will set the mu M.

このように更新量を調整する係数を設定することで、実施例4と同様の効果が得られる他に、次のような効果も得られる。受話信号x(n)の音量が小さくSN比が悪い場合には、更新量w^(n)が小さくなるので反響消去量が低下しないようにできる。また、受話信号x(n)の音量が大きくSN比がよい場合には、更新量w^(n)が大きくなるので収束速度を速くできる。   In this way, by setting the coefficient for adjusting the update amount, the same effect as in the fourth embodiment can be obtained, and the following effect can also be obtained. When the volume of the received signal x (n) is small and the S / N ratio is bad, the update amount w ^ (n) is small, so that the echo cancellation amount can be prevented from decreasing. Further, when the volume of the received signal x (n) is large and the SN ratio is good, the update amount w ^ (n) becomes large, so that the convergence speed can be increased.

[変形例2]
実施例4では、係数計算部430がM個の重み係数β,…,βを求める方法として式(24)の例を示した。しかし、この式に限定する必要はない。例えば、式(24)の代わりに、
β=f(x^(n),v,v,…,v) (30)
ただし、
[Modification 2]
In Example 4, the example of Formula (24) was shown as a method in which the coefficient calculation part 430 calculates | requires M weighting coefficient (beta) 1 , ..., (beta) M. However, it is not necessary to limit to this formula. For example, instead of equation (24),
β i = f i (x ^ (n), v 1 , v 2 ,..., v M ) (30)
However,

Figure 0005149833
Figure 0005149833

を用いてもよい。この式の場合、重みは正規分布を用いて決めることになる。なお、σは正規分布の標準偏差である。 May be used. In the case of this equation, the weight is determined using a normal distribution. Here, σ is a standard deviation of normal distribution.

図12に実施例5の反響消去装置の機能構成例を、図13に実施例5の反響消去装置の処理フロー例を示す。実施例5の反響消去装置500は、固定フィルタ部510が追加された点と、重み付き加算部561が固定フィルタ部510からの出力も加算する点が反響消去装置400と異なる。固定フィルタ部510は、受話信号x(n)をフィルタリングし、固定擬似エコー信号z(n)を出力する(S510)。例えば、固定フィルタ部510のタップ長をL’(ただし、L’は2以上の整数)、固定フィルタ係数h^を
^=[h1,f,h2,f,…,hL’,f] (32)
とすると、固定フィルタ部510から出力される固定擬似エコー信号z(n)は、
FIG. 12 shows a functional configuration example of the echo canceling apparatus of the fifth embodiment, and FIG. 13 shows a processing flow example of the echo canceling apparatus of the fifth embodiment. The echo canceling apparatus 500 according to the fifth embodiment is different from the echo canceling apparatus 400 in that a fixed filter unit 510 is added and a weighted adding unit 561 also adds an output from the fixed filter unit 510. The fixed filter unit 510 filters the received signal x (n) and outputs a fixed pseudo echo signal z f (n) (S510). For example, the tap length of the fixed filter unit 510 is set to L ′ (where L ′ is an integer equal to or greater than 2), and the fixed filter coefficient h f ^ is set to h f ^ = [h 1, f , h 2, f ,. ', F ] T (32)
Then, the fixed pseudo echo signal z f (n) output from the fixed filter unit 510 is

Figure 0005149833
Figure 0005149833

となる。なお、固定フィルタ係数h^は、無響室であらかじめ測定しておいたある音量でのインパルス応答などを用いればよい。
重み付き加算部561は、適応フィルタ部410−1,…,410−Mが出力した個別擬似エコー信号z(n),…,z(n)と固定擬似エコー信号z(n)とを重み付け加算した信号を、擬似エコー信号z(n)とする(S561)。例えば、重み付け加算部561は、係数乗算部440−1,…,440−Mと加算部560で構成すればよい。係数乗算部440−mは、入力された個別擬似エコー信号z(n)と重み係数βとを乗算し、β(n)を出力する(S440−1,…,S440−M)。これらの乗算結果と固定擬似エコー信号z(n)を、加算部560が加算することで、擬似エコー信号z(n)を
It becomes. Note that the fixed filter coefficient h f ^ may be an impulse response at a certain volume measured in advance in an anechoic chamber.
The weighted addition unit 561 includes individual pseudo echo signals z 1 (n),..., Z M (n) and fixed pseudo echo signals z f (n) output from the adaptive filter units 410-1,. Is a pseudo echo signal z (n) (S561). For example, the weighting addition unit 561 may be configured by coefficient multiplication units 440-1,..., 440-M and an addition unit 560. The coefficient multiplier 440-m multiplies the input individual pseudo echo signal z m (n) and the weight coefficient β m and outputs β m z m (n) (S440-1,..., S440-M). ). The addition unit 560 adds these multiplication results and the fixed pseudo echo signal z f (n), thereby obtaining the pseudo echo signal z (n).

Figure 0005149833
Figure 0005149833

のように求める(S560)。その他の処理フローは実施例4と同じである。また、実施例4変形例1のように、更新量を調整する係数がそれぞれ異なる適応フィルタ部410’−1,…,410’−Mを用いてもよい。 (S560). Other processing flows are the same as those in the fourth embodiment. Further, as in the fourth modification example of the fourth embodiment, adaptive filter units 410'-1, ..., 410'-M having different coefficients for adjusting the update amount may be used.

反響消去装置500は固定フィルタ部510を備えているので、M個の適応フィルタ部410−1,…,410−Mは、固定フィルタ係数を求めたときの条件との差によって生じる反響の変化分を推定することになる。例えば、固定フィルタ係数が、無響室であらかじめ測定しておいたある音量でのインパルス応答の場合、実際の部屋での反響の違いと受話信号の音量の違いによるインパルス応答の変化のみを推定することになる。このインパルス応答の変化が、固定フィルタ部510のタップ長L’よりも短いタップ数で表現できる場合、適応フィルタ部410−mのタップ数LをL’よりも小さくできる。したがって、反響消去装置500では、実施例4や実施例4変形例1と同様の効果が得られる他に、適応フィルタ係数h^(n)の要素の数を少なくでき、演算量を減らすことも期待できる。 Since the echo canceling apparatus 500 includes the fixed filter unit 510, the M adaptive filter units 410-1,..., 410-M have the amount of change in the echo caused by the difference from the condition when the fixed filter coefficient is obtained. Will be estimated. For example, if the fixed filter coefficient is an impulse response at a certain volume measured in an anechoic room in advance, only the change in the impulse response due to the difference in the echo in the actual room and the volume of the received signal is estimated. It will be. When the change in the impulse response can be expressed by a tap number shorter than the tap length L ′ of the fixed filter unit 510, the tap number L of the adaptive filter unit 410-m can be made smaller than L ′. Therefore, in the echo canceling apparatus 500, in addition to the same effects as those of the fourth embodiment and the fourth modification of the first embodiment, the number of elements of the adaptive filter coefficient h m ^ (n) can be reduced and the amount of calculation can be reduced. Can also be expected.

図14に実施例6の反響消去装置の機能構成例を、図15に実施例6の反響消去装置の処理フロー例を示す。実施例6の反響消去装置600は、遅延部370、適応フィルタ部610が追加された点と、重み付き加算部661が適応フィルタ部610の出力も加算する点が反響消去装置400と異なる。また、適応フィルタ部610の機能構成例は図10に示す。   FIG. 14 shows a functional configuration example of the echo canceling apparatus according to the sixth embodiment, and FIG. 15 shows a processing flow example of the echo canceling apparatus according to the sixth embodiment. The echo canceling apparatus 600 of the sixth embodiment is different from the echo canceling apparatus 400 in that a delay unit 370 and an adaptive filter unit 610 are added, and a weighted adding unit 661 also adds the output of the adaptive filter unit 610. An example of the functional configuration of the adaptive filter unit 610 is shown in FIG.

遅延部370は、受話信号x(n)をあらかじめ定めた時間Dだけ遅延させて、遅延信号x(n−D)を出力する(S370)。適応フィルタ部610は、フィルタ手段611(616)と更新量計算手段612(617)とを有する。適応フィルタ部610のフィルタ手段611は、遅延信号x(n−D)をフィルタリングすることで遅延擬似エコー信号z(n)を生成する(S611)。例えば、適応フィルタ部610のタップ長をL”(ただし、L”は2以上の整数)、適応フィルタ係数h^を
^(n)=[h1,0(n),h2,0(n),…,hL”,0(n)] (35)
とすると、適応フィルタ部610から出力される遅延擬似エコー信号z(n)は、
The delay unit 370 delays the received signal x (n) by a predetermined time D and outputs a delayed signal x (n−D) (S370). The adaptive filter unit 610 includes filter means 611 (616) and update amount calculation means 612 (617). The filter unit 611 of the adaptive filter unit 610 generates the delayed pseudo echo signal z d (n) by filtering the delayed signal x (n−D) (S611). For example, the tap length of the adaptive filter unit 610 is L ″ (where L ″ is an integer equal to or greater than 2), and the adaptive filter coefficient h 0 ^ is set to h 0 ^ (n) = [h 1,0 (n), h 2, 0 (n),..., H L ″, 0 (n)] T (35)
Then, the delayed pseudo echo signal z d (n) output from the adaptive filter unit 610 is

Figure 0005149833
Figure 0005149833

となる。
重み付き加算部661は、適応フィルタ部410−1,…,410−Mが出力した個別擬似エコー信号z(n),…,z(n)と遅延擬似エコー信号z(n)とを重み付き加算した信号を、擬似エコー信号z(n)とする(S361)。例えば、重み付け加算部661は、係数乗算部440−1,…,440−Mと加算部660で構成すればよい。係数乗算部440−mは、入力された個別擬似エコー信号z(n)と重み係数βとを乗算し、β(n)を出力する(S440−1,…,S440−M)。これらの乗算結果と遅延擬似エコー信号z(n)とを、加算部660が加算することで、擬似エコー信号z(n)を
It becomes.
The weighted addition unit 661 includes individual pseudo echo signals z 1 (n),..., Z M (n) and delayed pseudo echo signals z d (n) output from the adaptive filter units 410-1,. A signal obtained by weighting and adding is set as a pseudo echo signal z (n) (S361). For example, the weighting addition unit 661 may be configured by coefficient multiplication units 440-1,..., 440-M and an addition unit 660. The coefficient multiplier 440-m multiplies the input individual pseudo echo signal z m (n) and the weight coefficient β m and outputs β m z m (n) (S440-1,..., S440-M). ). The addition unit 660 adds these multiplication results and the delayed pseudo echo signal z d (n), thereby obtaining the pseudo echo signal z (n).

Figure 0005149833
Figure 0005149833

のように求める(S460)。
適応フィルタ部610の更新量計算手段612は、受話信号x(n)と送話信号e(n)からフィルタ手段611を更新する(S612)。具体的には、更新量計算手段312は、更新量w^(n)を
(S460).
The update amount calculation unit 612 of the adaptive filter unit 610 updates the filter unit 611 from the received signal x (n) and the transmitted signal e (n) (S612). Specifically, the update amount calculation means 312 calculates the update amount w ^ (n).

Figure 0005149833
Figure 0005149833

のように求め、フィルタ手段111−mの次の時刻n+1の適応フィルタ係数h^(n+1)を、
^(n+1)=h^(n)+w^(n) (39)
のように毎時刻更新する。その他の処理フローは実施例4と同じである。また、実施例4変形例1のように、更新量を調整する係数がそれぞれ異なる適応フィルタ部410’−1,…,410’−Mを用いてもよい。
And the adaptive filter coefficient h 0 ^ (n + 1) at the next time n + 1 of the filter means 111-m is
h 0 ^ (n + 1) = h 0 ^ (n) + w ^ (n) (39)
Update every hour as follows. Other processing flows are the same as those in the fourth embodiment. Further, as in the fourth modification example of the fourth embodiment, adaptive filter units 410′-1,..., 410′-M having different coefficients for adjusting the update amount may be used.

反響消去装置600では、L+L”が実質的なタップ長となる。系の非線形性の影響がフィルタのはじめのL個までにしか含まれていない場合、非線形性に対応するためにM個備えられている適応フィルタ部410−mのタップ長Lを短くできる。そして、音量に依存しない線形な範囲を適応フィルタ部610で対応すればよい。このようにLとL”とを定めれば、反響消去の性能は反響消去装置400と同等にできる。さらに、適応フィルタ係数を記録するメモリは、M×L+L”必要であるが、Lを短くできるので全体としてはメモリを節約できる。例えば、ΔLだけ短くできた場合、反響消去装置400の場合よりも、M×ΔL−L”だけメモリを節約できる。   In the echo canceller 600, L + L ″ is a substantial tap length. When the influence of the nonlinearity of the system is included only in the first L of the filter, M is provided to cope with the nonlinearity. The tap length L of the adaptive filter unit 410-m can be shortened, and a linear range that does not depend on the sound volume can be dealt with by the adaptive filter unit 610. If L and L ″ are determined in this way, the echo will be reflected. The erasing performance can be equivalent to that of the echo canceling device 400. Further, the memory for recording the adaptive filter coefficient requires M × L + L ″, but L can be shortened, so that the memory can be saved as a whole. , M × ΔL−L ″ can save memory.

図16に、コンピュータの機能構成例を示す。なお、本発明の反響消去装置は、コンピュータ2000の記録部2020に、本発明の各構成部としてコンピュータ2000を動作させるプログラムを読み込ませ、処理部2010、入力部2030、出力部2040などを動作させることで実現できる。また、コンピュータに読み込ませる方法としては、プログラムをコンピュータ読み取り可能な記録媒体に記録しておき、記録媒体からコンピュータに読み込ませる方法、サーバ等に記録されたプログラムを、電気通信回線等を通じてコンピュータに読み込ませる方法などがある。   FIG. 16 shows a functional configuration example of a computer. Note that the echo canceling apparatus of the present invention causes the recording unit 2020 of the computer 2000 to read a program for operating the computer 2000 as each component of the present invention and operate the processing unit 2010, the input unit 2030, the output unit 2040, and the like. This can be achieved. In addition, as a method of causing the computer to read, the program is recorded on a computer-readable recording medium, and the program recorded on the server or the like is read into the computer through a telecommunication line or the like. There is a method to make it.

効果が生じる理由の分析
次に、本発明の効果を確認する。非線形歪みがインパルス応答に及ぼす影響を調べるために、振幅の異なる複数のM系列を用いてインパルス応答の測定を行った。図17に、M系列を用いて測定した5つのインパルス応答を示す。図18は、図17の点線で囲んだ部分を拡大した図である。振幅の小さいM系列で測定したものから順にインパルス応答1〜5と表記している。各インパルス応答はほぼ同じ形をしているが、ピーク値が異なっていることが分かる。
Analysis of the reasons for effect occurs Next, to confirm the effects of the present invention. In order to investigate the influence of nonlinear distortion on the impulse response, the impulse response was measured using a plurality of M series having different amplitudes. FIG. 17 shows five impulse responses measured using the M series. FIG. 18 is an enlarged view of a portion surrounded by a dotted line in FIG. The impulse responses are expressed as impulse responses 1 to 5 in order from the one measured with the M series having a small amplitude. It can be seen that each impulse response has almost the same shape, but the peak values are different.

次に、各インパルス応答のピーク値の変動に着目する。まず、測定信号の振幅が一番小さく、非線形歪みの影響が一番少ないと考えられるインパルス応答1を基準とする。ただし、インパルス応答1も雑音の影響は受けている。そして、インパルス応答1とその他のインパルス応答の振幅波形の差分を、インパルス応答のはじめの10個のピーク(12サンプル目から32サンプル目の間)において算出する。その結果を図19に示す。図19の横軸はM系列の振幅、縦軸はインパルス応答の差分である。この図によると、インパルス応答の差分は、多くの場合、入力であるM系列の振幅に応じて単調に増加または減少することが分かる。   Next, attention is paid to the fluctuation of the peak value of each impulse response. First, the impulse response 1 that is considered to have the smallest amplitude of the measurement signal and the least influence of nonlinear distortion is used as a reference. However, the impulse response 1 is also affected by noise. Then, the difference between the amplitude waveforms of the impulse response 1 and the other impulse responses is calculated at the first 10 peaks (between the 12th and 32nd samples) of the impulse response. The result is shown in FIG. In FIG. 19, the horizontal axis represents the amplitude of the M series, and the vertical axis represents the difference in impulse response. According to this figure, it can be seen that in many cases, the difference in impulse response monotonously increases or decreases according to the amplitude of the input M-sequence.

このように、系の非線形性はインパルス応答に影響を及ぼすだけでなく、入力信号の大きさによってインパルス応答が規則的に変化することが分かった。これが、入力信号の大きさに応じてインパルス応答を別々に推定することが有効な理由である。   Thus, it was found that the nonlinearity of the system not only affects the impulse response, but also the impulse response changes regularly depending on the magnitude of the input signal. This is the reason why it is effective to separately estimate the impulse response according to the magnitude of the input signal.

実験結果
筺体内に小型のスピーカ(40mm×20mm)とマイクロホンを設置したものを用い、簡易無響室で測定したデータに対する評価を図20に示す。音声データは8kHzサンプリング、40秒の女声(読み上げ音声)を用いた。このうち最初の10秒間は従来通り単一の適応フィルタ部を学習し、そこまでの推定値をすべての適応フィルタ部の初期値とした。そして、反響消去装置100と反響消去装置300の構成で10秒後から40秒後まで学習した。平均エコー消去量は、20秒後から40秒後の音声が存在する区間で算出した。また、20秒後に適応フィルタの更新を止めた場合についても、同様に平均エコー消去量を計算した。実験に用いたパラメータは、タップ長L=256、適応フィルタ部の数M=13、ステップサイズμ=0.2である。
Experimental Results FIG. 20 shows an evaluation of data measured in a simple anechoic chamber using a small speaker (40 mm × 20 mm) and a microphone installed in the housing. As voice data, 8 kHz sampling and 40 seconds of female voice (reading voice) were used. Of these, the first adaptive filter unit was learned as before for the first 10 seconds, and the estimated values up to that point were used as initial values for all the adaptive filter units. Then, learning was performed from 10 seconds to 40 seconds later with the configuration of the echo canceling apparatus 100 and the echo canceling apparatus 300. The average echo cancellation amount was calculated in a section in which there was a voice after 20 seconds to 40 seconds later. Also, when the updating of the adaptive filter was stopped after 20 seconds, the average echo cancellation amount was similarly calculated. Parameters used in the experiment are tap length L = 256, number of adaptive filter units M = 13, and step size μ = 0.2.

図20から、図1に示した従来の反響消去装置900よりも、反響消去装置100や反響消去装置300の方がエコー消去量で上回っていることが分かる。   From FIG. 20, it can be seen that the echo canceling apparatus 100 and the echo canceling apparatus 300 exceed the echo canceling amount in comparison with the conventional echo canceling apparatus 900 shown in FIG.

100、200、300 反響消去装置 110、310 適応フィルタ部
111、116、311 フィルタ手段 112、117、312 更新量計算手段
113、115 周波数変換手段 114 周波数逆変換手段
130 音量判定部 140 受話信号切替部
150 送話信号切替部 210 固定フィルタ部
260、360 加算部 370 遅延部
400、500、600 反響消去装置 410、610 適応フィルタ部
411、611 フィルタ手段 412、417、612 更新量計算手段
430 係数計算部 440 係数乗算部
460、560、660 加算部 461、561、661 重み付き加算部
510 固定フィルタ部 900 反響消去装置
910 適応フィルタ部 911 フィルタ手段
912 更新量計算手段 920 減算部
100, 200, 300 Echo canceller 110, 310 Adaptive filter unit 111, 116, 311 Filter unit 112, 117, 312 Update amount calculation unit 113, 115 Frequency conversion unit 114 Frequency inverse conversion unit 130 Volume determination unit 140 Received signal switching unit 150 Transmission signal switching unit 210 Fixed filter unit 260, 360 Adder unit 370 Delay unit 400, 500, 600 Echo canceling device 410, 610 Adaptive filter unit 411, 611 Filter unit 412, 417, 612 Update amount calculation unit 430 Coefficient calculation unit 440 Coefficient multiplication unit 460, 560, 660 Addition unit 461, 561, 661 Weighted addition unit 510 Fixed filter unit 900 Echo canceling device 910 Adaptive filter unit 911 Filter unit 912 Update amount calculation unit 920 Subtraction unit

Claims (9)

受話信号と収音信号が入力され、送話信号を出力する反響消去装置であって、
前記受話信号の音量があらかじめ定めたM個(ただし、Mは2以上の整数)の音量範囲の中のどの音量範囲に属すかを示す音量情報を出力する音量判定部と、
前記受話信号をフィルタリングすることで擬似エコー信号を生成するフィルタ手段と、前記受話信号と前記送話信号から前記フィルタ手段を更新する更新量計算手段とをそれぞれ有し、前記M個の音量情報のそれぞれに対応するM個の適応フィルタ部と、
記受話信号が入力される適応フィルタ部を、前記音量判定部が出力した音量情報と対応する適応フィルタ部に切り替える受話信号切替部と、
記送話信号が入力される適応フィルタ部を、前記音量判定部が出力した音量情報と対応する適応フィルタ部に切り替える送話信号切替部と、
前記収音信号から前記擬似エコー信号を減算して送話信号を求める減算部と
を備える反響消去装置。
An echo canceller that receives a received signal and a collected sound signal and outputs a transmitted signal,
M pieces of sound volume of the received signal is predetermined (although, M is an integer of 2 or more) and volume determination section that outputs what belongs to the volume range shown to volume information in the volume range,
Said filter means for generating a pseudo echo signal by filtering the received signal, said from the receiving signal and the transmission signal and the updating amount calculating means for updating the filter means and their respective closed, the M corresponding to each of the volume information, and M of the adaptive filter section,
An adaptive filter unit before Symbol received signal is input, and the reception signal switching unit for switching the adaptive filter section corresponding to the sound information volume determination unit is output,
Before Symbol adaptive filter unit transmission signal is inputted, it switches to adaptive filter section corresponding to the sound information volume judging unit has output transmission signal switching section,
An echo canceller comprising: a subtracting unit that subtracts the pseudo echo signal from the collected sound signal to obtain a transmission signal.
請求項1記載の反響消去装置であって、
さらに
前記受話信号をフィルタリングし、固定擬似エコー信号を生成する固定フィルタ部と、
前記擬似エコー信号と前記固定擬似エコー信号とを加算した信号を、新たな擬似エコー信号とする加算部と
を備えることを特徴とする反響消去装置。
The echo canceling device according to claim 1,
Further, a fixed filter unit that filters the received signal and generates a fixed pseudo echo signal;
An echo canceller, comprising: an adder that sets a signal obtained by adding the pseudo echo signal and the fixed pseudo echo signal as a new pseudo echo signal.
請求項1記載の反響消去装置であって、
さらに、
前記受話信号をあらかじめ定めた時間分遅延させて、遅延信号を出力する遅延部と、
前記遅延信号をフィルタリングすることで遅延擬似エコー信号を生成する第2フィルタ手段と、前記受話信号と前記送話信号から前記第2フィルタ手段を更新する第2更新量計算手段とを有する第2適応フィルタ部と、
前記擬似エコー信号と前記遅延擬似エコー信号とを加算した信号を、新たな擬似エコー信号とする加算部と
を備えることを特徴とする反響消去装置。
The echo canceling device according to claim 1,
further,
A delay unit that delays the reception signal by a predetermined time and outputs a delay signal;
A second adaptive unit comprising: second filter means for generating a delayed pseudo echo signal by filtering the delayed signal; and second update amount calculating means for updating the second filter means from the received signal and the transmitted signal. A filter section;
An echo canceller, comprising: an adder that sets a signal obtained by adding the pseudo echo signal and the delayed pseudo echo signal as a new pseudo echo signal.
請求項1から3のいずれかに記載の反響消去装置であって、
主にフィルタリングする受話信号の音量が大きい適応フィルタ部ほど、当該適応フィルタ部の更新量計算手段が求める更新の量を調整する係数が大きい値である
ことを特徴とする反響消去装置。
A echo canceller according to claim 1 or et 3,
An echo canceling apparatus characterized in that an adaptive filter unit having a larger volume of an incoming signal to be filtered mainly has a larger coefficient for adjusting an update amount obtained by an update amount calculation means of the adaptive filter unit.
受話信号と収音信号が入力され、適応フィルタを用いて擬似エコー信号を推定し、送話信号を出力する反響消去方法であって、
前記受話信号の音量があらかじめ定めたM個(ただし、Mは2以上の整数)の音量範囲の中のどの音量範囲に属すかを判定し、該当する音量範囲を示す音量情報を出力する音量判定ステップと、
前記音量判定ステップが出力した音量情報と対応する適応フィルタに前記受話信号を入力する受話信号切替ステップと、
前記音量判定ステップが出力した音量情報と対応する適応フィルタに前記送話信号を入力する送話信号切替ステップと、
前記受話信号を適応フィルタでフィルタリングすることで擬似エコー信号を生成するフィルタサブステップと、前記受話信号と前記送話信号から前記フィルタサブステップの適応フィルタを更新する更新量計算サブステップとを、それぞれ有するM個の適応フィルタステップと、
前記収音信号から前記擬似エコー信号を減算して送話信号を求める減算ステップと
を有する反響消去方法。
An echo cancellation method for receiving a received signal and a collected sound signal, estimating a pseudo echo signal using an adaptive filter, and outputting a transmitted signal,
Volume determination for determining which volume range the volume of the received signal belongs to a predetermined volume range of M (where M is an integer of 2 or more), and outputting volume information indicating the corresponding volume range Steps,
And the reception signal switching step of inputting pre Symbol received signal to the adaptive filter corresponding to the volume information that the volume determination step is output,
And transmission signal switching step of inputting the pre Symbol transmission signal to the adaptive filter to the volume determination step corresponds to the volume information output,
A filter sub-step for generating a pseudo echo signal by filtering the received signal with an adaptive filter, and an update amount calculating sub-step for updating an adaptive filter of the filter sub-step from the received signal and the transmitted signal, respectively. M adaptive filter steps having:
A subtraction step of subtracting the pseudo echo signal from the collected sound signal to obtain a transmission signal.
請求項5記載の反響消去方法であって、
さらに
前記受話信号をフィルタリングし、固定擬似エコー信号を生成する固定フィルタステップと、
前記擬似エコー信号と前記固定擬似エコー信号とを加算した信号を、前記減算ステップで用いる擬似エコー信号として出力する加算ステップと
を有することを特徴とする反響消去方法。
A claim 5 Symbol mounting the echo cancellation method,
A fixed filter step of filtering the received signal to generate a fixed pseudo echo signal;
An echo canceling method comprising: an addition step of outputting a signal obtained by adding the pseudo echo signal and the fixed pseudo echo signal as a pseudo echo signal used in the subtraction step.
請求項5記載の反響消去方法であって、
さらに、
前記受話信号をあらかじめ定めた時間分遅延させて、遅延信号を出力する遅延ステップと、
前記遅延信号を適応フィルタでフィルタリングすることで遅延擬似エコー信号を生成する第2フィルタサブステップと、前記受話信号と前記送話信号から前記第2フィルタサブステップの適応フィルタを更新する第2更新量計算サブステップとを有する第2適応フィルタステップと、
前記擬似エコー信号と前記遅延擬似エコー信号とを加算した信号を、前記減算ステップで用いる擬似エコー信号として出力する加算ステップと
を有することを特徴とする反響消去方法。
A claim 5 Symbol mounting the echo cancellation method,
further,
A delay step of delaying the received signal by a predetermined time and outputting a delayed signal;
A second filter substep for generating a delayed pseudo echo signal by filtering the delayed signal with an adaptive filter; and a second update amount for updating the adaptive filter of the second filter substep from the received signal and the transmitted signal. A second adaptive filter step having a calculation sub-step;
An echo canceling method comprising: an addition step of outputting a signal obtained by adding the pseudo echo signal and the delayed pseudo echo signal as a pseudo echo signal used in the subtraction step.
請求項5から7のいずれかに記載の反響消去方法であって、
主にフィルタリングする受話信号の音量が大きい適応フィルタステップほど、当該適応フィルタステップの更新量計算サブステップが求める更新の量を調整する係数が大きい値である
ことを特徴とする反響消去方法。
The echo cancellation method according to any one of claims 5 to 7 ,
An echo canceling method characterized in that the coefficient for adjusting the amount of update obtained by the update amount calculation sub-step of the adaptive filter step is a larger value as the volume of the received signal to be filtered mainly increases.
請求項1から4のいずれかに記載の反響消去装置としてコンピュータを動作させる反響消去プログラム。 Echo cancellation program causing a computer to operate as echo canceller according to claim 1 or al 4.
JP2009040644A 2009-02-24 2009-02-24 Echo canceling device, echo canceling method, echo canceling program Expired - Fee Related JP5149833B2 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP2009040644A JP5149833B2 (en) 2009-02-24 2009-02-24 Echo canceling device, echo canceling method, echo canceling program

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
JP2009040644A JP5149833B2 (en) 2009-02-24 2009-02-24 Echo canceling device, echo canceling method, echo canceling program

Related Child Applications (1)

Application Number Title Priority Date Filing Date
JP2012180010A Division JP5406966B2 (en) 2012-08-15 2012-08-15 Echo canceling device, echo canceling method, echo canceling program

Publications (2)

Publication Number Publication Date
JP2010199809A JP2010199809A (en) 2010-09-09
JP5149833B2 true JP5149833B2 (en) 2013-02-20

Family

ID=42824095

Family Applications (1)

Application Number Title Priority Date Filing Date
JP2009040644A Expired - Fee Related JP5149833B2 (en) 2009-02-24 2009-02-24 Echo canceling device, echo canceling method, echo canceling program

Country Status (1)

Country Link
JP (1) JP5149833B2 (en)

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2002016691A (en) * 2000-06-29 2002-01-18 Toshiba Corp Echo canceller
JP4655905B2 (en) * 2005-11-25 2011-03-23 ヤマハ株式会社 Regression sound removal device
JP4702371B2 (en) * 2005-10-26 2011-06-15 日本電気株式会社 Echo suppression method and apparatus

Also Published As

Publication number Publication date
JP2010199809A (en) 2010-09-09

Similar Documents

Publication Publication Date Title
CN107636758B (en) Acoustic echo cancellation system and method
AU694622B2 (en) Apparatus and method for canceling acoustic echoes including non-linear distortions in loudspeaker telephones
Guo et al. Novel acoustic feedback cancellation approaches in hearing aid applications using probe noise and probe noise enhancement
AU756511B2 (en) Signal noise reduction by spectral subtraction using linear convolution and causal filtering
JP5038143B2 (en) Echo cancellation
JP5751110B2 (en) Reverberation suppression apparatus, reverberation suppression method, and reverberation suppression program
WO2001056328A1 (en) System and method for dual microphone signal noise reduction using spectral subtraction
WO2015086229A9 (en) Echo cancellation
EP2410763A1 (en) Howling canceller
US9667803B2 (en) Nonlinear acoustic echo cancellation based on transducer impedance
TW200844980A (en) Apparatus with an acoustic echo cancellation system and acoustic echo cancellation method
JP2013507804A (en) Time domain / frequency domain integrated echo cancellation apparatus and method
JP2004349806A (en) Multichannel acoustic echo canceling method, apparatus thereof, program thereof, and recording medium thereof
JP5662232B2 (en) Echo canceling apparatus, method and program
JP5406966B2 (en) Echo canceling device, echo canceling method, echo canceling program
JP3787088B2 (en) Acoustic echo cancellation method, apparatus, and acoustic echo cancellation program
JP3756828B2 (en) Reverberation elimination method, apparatus for implementing this method, program, and recording medium therefor
JP2005051744A (en) Speech communication apparatus
KR101442700B1 (en) Noise cancellation method and apparatus using independent component analysis for headphones
JP5149833B2 (en) Echo canceling device, echo canceling method, echo canceling program
JP4041770B2 (en) Acoustic echo cancellation method, apparatus, program, and recording medium
JP2002223182A (en) Echo canceling method, its device, its program and its recording medium
CN112863532A (en) Echo suppressing device, echo suppressing method, and storage medium
JP6143702B2 (en) Echo canceling apparatus, method and program
JP3917116B2 (en) Echo canceling apparatus, method, echo canceling program, and recording medium recording the program

Legal Events

Date Code Title Description
A621 Written request for application examination

Free format text: JAPANESE INTERMEDIATE CODE: A621

Effective date: 20110118

RD03 Notification of appointment of power of attorney

Free format text: JAPANESE INTERMEDIATE CODE: A7423

Effective date: 20110715

A977 Report on retrieval

Free format text: JAPANESE INTERMEDIATE CODE: A971007

Effective date: 20120611

A131 Notification of reasons for refusal

Free format text: JAPANESE INTERMEDIATE CODE: A131

Effective date: 20120619

A521 Written amendment

Free format text: JAPANESE INTERMEDIATE CODE: A523

Effective date: 20120815

TRDD Decision of grant or rejection written
A01 Written decision to grant a patent or to grant a registration (utility model)

Free format text: JAPANESE INTERMEDIATE CODE: A01

Effective date: 20121120

A61 First payment of annual fees (during grant procedure)

Free format text: JAPANESE INTERMEDIATE CODE: A61

Effective date: 20121130

R150 Certificate of patent or registration of utility model

Ref document number: 5149833

Country of ref document: JP

Free format text: JAPANESE INTERMEDIATE CODE: R150

Free format text: JAPANESE INTERMEDIATE CODE: R150

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20151207

Year of fee payment: 3

S531 Written request for registration of change of domicile

Free format text: JAPANESE INTERMEDIATE CODE: R313531

R350 Written notification of registration of transfer

Free format text: JAPANESE INTERMEDIATE CODE: R350

LAPS Cancellation because of no payment of annual fees