JPS63195699A - Pitch extraction - Google Patents
Pitch extractionInfo
- Publication number
- JPS63195699A JPS63195699A JP2883687A JP2883687A JPS63195699A JP S63195699 A JPS63195699 A JP S63195699A JP 2883687 A JP2883687 A JP 2883687A JP 2883687 A JP2883687 A JP 2883687A JP S63195699 A JPS63195699 A JP S63195699A
- Authority
- JP
- Japan
- Prior art keywords
- sample
- absolute value
- pitch
- pitch period
- samples
- 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.)
- Granted
Links
- 238000000605 extraction Methods 0.000 title description 13
- 238000000034 method Methods 0.000 claims description 6
- 238000001514 detection method Methods 0.000 description 8
- 230000015572 biosynthetic process Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000000737 periodic effect Effects 0.000 description 2
- 238000003786 synthesis reaction Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
Landscapes
- Working-Up Tar And Pitch (AREA)
- Fats And Perfumes (AREA)
- Steroid Compounds (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
- Electrophonic Musical Instruments (AREA)
Abstract
(57)【要約】本公報は電子出願前の出願データであるた
め要約のデータは記録されません。(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.
Description
【発明の詳細な説明】
産業上の利用分野
本発明は音声合成・音声認識等の音声処理に用いること
のできる、音声波形のピッチ抽出方法に関するものであ
る。DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention relates to a method for extracting the pitch of a speech waveform, which can be used in speech processing such as speech synthesis and speech recognition.
従来の技術
近年、コンピュータの発達と共に、音声合成・音声認識
等の音声処理技術の開発が急速に進められてきている。BACKGROUND OF THE INVENTION In recent years, along with the development of computers, the development of speech processing technologies such as speech synthesis and speech recognition has progressed rapidly.
なかでもピッチ抽出は音声処理を行なう際には不可欠な
技術である。Among these, pitch extraction is an indispensable technique when performing audio processing.
以下、上述した従来のピッチ抽出方法の一例について説
明する。関数x (n)が周期Pをもつ真の周期関数な
らば差数列
d(n)= xln) −x (n −k)が、k=0
.±P、±2P、・・・・・・で零になるということを
利用する。実際の音声では零にならないが、有声音では
かなり小さな値をとる。An example of the conventional pitch extraction method described above will be described below. If the function x (n) is a true periodic function with period P, then the difference sequence d (n) = xln) -x (n - k) becomes k = 0
.. The fact that it becomes zero at ±P, ±2P, ... is utilized. Although it does not become zero in actual speech, it takes a considerably small value in voiced speech.
d (n)の短時間平均振幅は、kの関数としてはkが
周期に近い値のときにいつでも小さくなるはずである。The short-term average amplitude of d(n) should be small as a function of k whenever k is close to the period.
短時間平均振幅差関数は次式で定義される。The short-time average amplitude difference function is defined by the following equation.
r、(k)= Σ l x(n十m)
WI (In!−x (n十m−k) w z (
m−k) lここで、Wr (n)、WI (n)
は窓関数である。もしもx (nlが窓の長さの範囲内
で周期的であれば、γ0(k)はに=P、2P、・・・
・・・で鋭い谷になるはずである。r, (k) = Σ l x (n0m)
WI (In!-x (n0m-k) w z (
m-k) lwhere, Wr (n), WI (n)
is a window function. If x (nl is periodic within the window length, γ0(k) is = P, 2P,...
...and it should become a sharp valley.
(例えばり、R,Rabiner/ R,W、5cha
fer著、鈴木久喜訳、「音声のディジタル信号処理(
上)」、157〜158ページ)。(For example, R, Rabiner/R, W, 5cha
fer, translated by Hisaki Suzuki, “Digital signal processing of audio (
(above), pp. 157-158).
発明が解決しようとする問題点
しかしながら上記のような方法では、kの値を変化させ
て、減算と加算と絶対値演算をくり返し行なわなければ
ならない。また、kの異なるγ7゛(k)の中で最も小
さな値を見つけなければならず、かつ最も小さなγ7(
k)を与えるkが、Pであって2P、3P、・・・・・
・ではないというチェックを行なわなければならない。Problems to be Solved by the Invention However, in the above method, the value of k must be changed and subtraction, addition, and absolute value operations must be repeated. Also, it is necessary to find the smallest value among γ7゛(k) with different k, and the smallest γ7(k)
k) is P and 2P, 3P, etc.
- It is necessary to check that this is not the case.
これらをパソコンを用いて処理する場合、時間がかかる
という問題点を有していた。When processing these using a personal computer, there was a problem in that it took a long time.
本発明は上記問題点に鑑み、パソコンを用いても短時間
でできる、簡便なピンチ抽出方法を提供するものである
。In view of the above problems, the present invention provides a simple pinch extraction method that can be performed in a short time using a personal computer.
問題点を解決するための手段
上記問題点を解決するために本発明のピッチ抽出方法は
、標本化された音声波形のうち、区間を設定して、区間
毎に絶対値の大きな標本数個を見つけるだけで、ピッチ
抽出を行なうものである。Means for Solving the Problems In order to solve the above problems, the pitch extraction method of the present invention sets sections of the sampled audio waveform and extracts several samples with large absolute values for each section. Just by finding the pitch, the pitch is extracted.
作用
本発明は上記した方法によって、標本を検出する区間お
よび、検出する標本の数を、区間毎に設定し、すべての
標本を検出したのち、最も絶対値和の大きな標本の位置
と検出した標本の位置の差の絶対値を算出してピッチ周
期候補とし、ピンチ周期候補の中で、最も値の小さいピ
ッチ周期候補を与える標本の絶対値と、最も絶対値の大
きな標本の絶対値の比率を算出し、比率があらかじめ定
めた範囲内にあれば、最も値の小さいピッチ周期候補を
所定の区間のピッチ周期と決定することとする。Effect The present invention uses the method described above to set the interval for detecting samples and the number of samples to be detected for each interval, and after detecting all the samples, calculate the position of the sample with the largest sum of absolute values and the detected sample. The absolute value of the difference between the positions of is calculated as a pitch period candidate, and among the pinch period candidates, the ratio of the absolute value of the sample that gives the pitch period candidate with the smallest value to the absolute value of the sample with the largest absolute value is calculated. If the ratio is within a predetermined range, the pitch period candidate with the smallest value is determined to be the pitch period of the predetermined section.
実施例
以下本発明の一実施例のピッチ抽出方法について、図面
を参照しながら説明する。EXAMPLE Hereinafter, a pitch extraction method according to an example of the present invention will be described with reference to the drawings.
第1図は本発明の一実施例におけるピッチ抽出方法を説
明するための音声波形図を示すものである。第1図にお
いて、1は最も絶対値の大きな標本とその位置を検出す
るためのあらかじめ定めた標本検出区間、2は1におけ
る最も絶対値の大きな標本、3は2を基準に定めた、標
本を検出するための標本検出区間、4は3における、2
と同符号の絶対値の大きな標本、5は2と4の位置の差
で、ピッチ周期候補である。FIG. 1 shows an audio waveform diagram for explaining a pitch extraction method in an embodiment of the present invention. In Figure 1, 1 is a predetermined sample detection interval for detecting the sample with the largest absolute value and its position, 2 is the sample with the largest absolute value in 1, and 3 is the sample determined based on 2. Sample detection interval for detection, 4 in 3, 2
The sample with the same sign and large absolute value, 5, is the difference between the positions of 2 and 4, and is a pitch period candidate.
第2図は本発明の一実施例におけるピッチ抽出方法を説
明するためのフローチャートを示すものである。FIG. 2 shows a flowchart for explaining a pitch extraction method in one embodiment of the present invention.
本発明のピッチ抽出方法について、以下第1図および第
2図を用いてその方法を説明する。The pitch extraction method of the present invention will be described below with reference to FIGS. 1 and 2.
まず、標本検出区間lは区間長20nlSeCであり、
この中に含まれる標本のうち、最も絶対値の大きな標本
2を求める。2の位置を基準として標本検出区間3を設
定する。3は始端が2の位置から時間軸の正方向に3
m5ecのところであり、区間長は18m5ecである
。次にこの区間で、2と同符号である標本の中から、絶
対値の大きな順に従って所定の数の標本を検出する。第
1図の場合は所定の数は1個であり、標本4を検出する
。4の位置と2の位置の差5を求め、ピッチ周期候補と
する。次に標本の絶対値を用いてピッチ周期候補のピッ
チとしての妥当性を検査する。標本2の絶対値をAt、
標本4の絶対値をA4とすると、A2とA4が条件
Ax/A4<k ; kは正の定数を満たせば、ピ
ッチ周期候補5をピッチ周期と決定し、上記条件を満た
さなければ、ピッチ周期をOと決定する。なお標本検出
区間3の始端および終端は2の位置から時間軸の負方向
に設定することも可能である。First, the sample detection interval l has an interval length of 20nlSeC,
Among the samples included therein, sample 2 with the largest absolute value is found. Specimen detection section 3 is set using position 2 as a reference. 3 means that the starting point is 3 in the positive direction of the time axis from the position of 2.
m5ec, and the section length is 18m5ec. Next, in this interval, a predetermined number of samples are detected from samples having the same sign as 2 in descending order of absolute value. In the case of FIG. 1, the predetermined number is one, and sample 4 is detected. The difference 5 between the position 4 and the position 2 is determined and used as a pitch period candidate. Next, the validity of the pitch period candidate as a pitch is checked using the absolute value of the sample. Let the absolute value of sample 2 be At,
If the absolute value of sample 4 is A4, then A2 and A4 satisfy the condition Ax/A4<k; k is a positive constant, then pitch period candidate 5 is determined to be the pitch period, and if the above condition is not satisfied, then pitch period candidate 5 is determined as the pitch period. is determined to be O. Note that the starting and ending ends of the sample detection section 3 can also be set in the negative direction of the time axis from the position 2.
以上のように本実施例によれば、標本の中から絶対値の
大きい標本を検出すればよいので、容易であり、しかも
検出されたピッチ周期候補の妥当性を検査しているので
、精度の高いピッチ抽出を行なうことができる。As described above, according to this embodiment, it is easy to detect a sample with a large absolute value from among the samples, and since the validity of the detected pitch period candidate is checked, the accuracy can be improved. High pitch extraction can be performed.
発明の効果
以上のように本発明は、音声波形の所定の区間における
絶対値の最も大きな標本とその位置を検出し、その位置
を基準とした区間の中から、先に検出した標本と同符号
でかつ絶対値の大きな標本をその位置と共に所定の数だ
け検出し、ピッチ周期候補の妥当性を、標本の絶対値を
用いて検査しているので、容易にしかも高い精度でピッ
チ抽出を行なうことができる。Effects of the Invention As described above, the present invention detects the sample with the largest absolute value and its position in a predetermined section of a speech waveform, and selects a sample with the same sign as the previously detected sample from the section based on that position. A predetermined number of samples with large and large absolute values are detected along with their positions, and the validity of pitch period candidates is checked using the absolute values of the samples, so pitch extraction can be performed easily and with high accuracy. Can be done.
第1図は本発明の一実施例におけるピッチ抽出方法を説
明するための音声波形図、第2図はピッチ抽出方法を示
すフローチャートである。
1・・・・・・標本検出区間、2・・・・・・標本、3
・・・・・・標本検出区間、4・・・・・・標本、5・
・・・・・ピッチ周期候補。
第2図FIG. 1 is an audio waveform diagram for explaining a pitch extraction method in an embodiment of the present invention, and FIG. 2 is a flowchart showing the pitch extraction method. 1... Sample detection interval, 2... Sample, 3
...Sample detection interval, 4...Sample, 5.
...Pitch period candidate. Figure 2
Claims (1)
た区間Tにおけるピッチ周期を抽出する方法であって、
上記区間Tにおいて絶対値の最も大きな標本S_0とそ
の位置t_0を検出し、上記t_0を基準にして、あら
たに区間T′を設定し、上記区間T′において、上記S
_0と同符号である標本の中から、絶対値の大きな順に
従って所定の数の標本S_i(i=1、2、・・・・・
・n)をその位置を1(i=1、2、・・・・・・n)
と共に検出し、所定の数すべての標本を検出したのち、
上記t_0と上記t_iの差の絶対値|t_0−t_i
|(i=1、2、・・・・・・n)を算出してピッチ周
期候補とし、最も小さい|t_0−t_i|を与える標
本S_iの絶対値|S_i|と、上記S_0の絶対値|
S_0|の比率を算出し、上記比率があらかじめ定めた
範囲内にあれば、上記最も値の小さい|t_0−t_i
|を上記区間Tのピッチ周期と決定し、上記比率があら
かじめ定めた範囲内になければ、上記区間Tのピッチ周
期を0と決定することを特徴とする、ピッチ抽出方法。A method for extracting a pitch period in a predetermined section T of a voice waveform sampled at a predetermined period, the method comprising:
The sample S_0 with the largest absolute value and its position t_0 in the above interval T are detected, a new interval T' is set based on the above t_0, and the above S_0 is detected in the above interval T'.
A predetermined number of samples S_i (i=1, 2,...
・n) and its position as 1 (i=1, 2,...n)
After detecting all the specified number of samples,
Absolute value of the difference between the above t_0 and the above t_i | t_0 - t_i
|(i=1, 2,...n) is calculated as a pitch period candidate, and the absolute value of the sample S_i that gives the smallest |t_0-t_i| |S_i| and the absolute value of the above S_0 |
Calculate the ratio of S_0|, and if the above ratio is within a predetermined range, the smallest value |t_0−t_i
| is determined as the pitch period of the section T, and if the ratio is not within a predetermined range, the pitch period of the section T is determined to be 0.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2883687A JPH079595B2 (en) | 1987-02-10 | 1987-02-10 | Pitch extraction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2883687A JPH079595B2 (en) | 1987-02-10 | 1987-02-10 | Pitch extraction method |
Publications (2)
Publication Number | Publication Date |
---|---|
JPS63195699A true JPS63195699A (en) | 1988-08-12 |
JPH079595B2 JPH079595B2 (en) | 1995-02-01 |
Family
ID=12259460
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
JP2883687A Expired - Lifetime JPH079595B2 (en) | 1987-02-10 | 1987-02-10 | Pitch extraction method |
Country Status (1)
Country | Link |
---|---|
JP (1) | JPH079595B2 (en) |
-
1987
- 1987-02-10 JP JP2883687A patent/JPH079595B2/en not_active Expired - Lifetime
Also Published As
Publication number | Publication date |
---|---|
JPH079595B2 (en) | 1995-02-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Barkani et al. | Amazigh speech recognition based on the Kaldi ASR toolkit | |
JPS63195699A (en) | Pitch extraction | |
JP4219539B2 (en) | Acoustic classification device | |
Zeng et al. | Modified AMDF pitch detection algorithm | |
Izzad et al. | Speech/non-speech detection in Malay language spontaneous speech | |
JP2574233B2 (en) | Pitch extraction method | |
JP2574234B2 (en) | Pitch extraction method | |
JP3031081B2 (en) | Voice recognition device | |
CN115910042B (en) | Method and device for identifying information type of formatted audio file | |
Khaing et al. | Automatic speech segmentation for myanmar language | |
Waghela et al. | SUV detection algorithm for speech signals | |
JP2583854B2 (en) | Voiced / unvoiced judgment method | |
JP4890792B2 (en) | Speech recognition method | |
Dersch | A decision logic for speech recognition | |
JPS622300A (en) | Voice pitch extractor | |
JPS58195895A (en) | Word voice recognition equipment | |
JPS60167000A (en) | Pitch extractor | |
JPS6069694A (en) | Segmentation of head consonant | |
JPH0398098A (en) | Voice recognition device | |
JPH01310400A (en) | Speech pitch extracting device | |
JPS58116595A (en) | Word voice recognition equipment | |
JPS61116397A (en) | Voice recognition | |
JPS6120099A (en) | Phoneme segmentation apparatus | |
JPS63155100A (en) | Max. point detection for self-correlation function given by formant | |
JPS6363919B2 (en) |