CN108735192B - System and method for evaluating piano playing tone quality by combining music - Google Patents
System and method for evaluating piano playing tone quality by combining music Download PDFInfo
- Publication number
- CN108735192B CN108735192B CN201810311753.2A CN201810311753A CN108735192B CN 108735192 B CN108735192 B CN 108735192B CN 201810311753 A CN201810311753 A CN 201810311753A CN 108735192 B CN108735192 B CN 108735192B
- Authority
- CN
- China
- Prior art keywords
- music
- piano
- signal
- evaluation
- recording device
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 31
- 238000011156 evaluation Methods 0.000 claims abstract description 83
- 238000013441 quality evaluation Methods 0.000 claims abstract description 57
- 230000005236 sound signal Effects 0.000 claims abstract description 34
- 238000004458 analytical method Methods 0.000 claims abstract description 31
- 230000008569 process Effects 0.000 claims abstract description 23
- 238000000605 extraction Methods 0.000 claims abstract description 18
- 238000010586 diagram Methods 0.000 claims abstract description 12
- 238000013528 artificial neural network Methods 0.000 claims description 41
- 238000012549 training Methods 0.000 claims description 23
- 238000012545 processing Methods 0.000 claims description 20
- 239000013598 vector Substances 0.000 claims description 20
- 230000006870 function Effects 0.000 claims description 18
- 238000012854 evaluation process Methods 0.000 claims description 7
- 238000002474 experimental method Methods 0.000 claims description 4
- 238000009432 framing Methods 0.000 claims description 4
- 238000010835 comparative analysis Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 2
- 238000011160 research Methods 0.000 description 9
- 238000011161 development Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000013139 quantization Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000008054 signal transmission Effects 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000003321 amplification Effects 0.000 description 1
- 230000019771 cognition Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- IXSZQYVWNJNRAL-UHFFFAOYSA-N etoxazole Chemical compound CCOC1=CC(C(C)(C)C)=CC=C1C1N=C(C=2C(=CC=CC=2F)F)OC1 IXSZQYVWNJNRAL-UHFFFAOYSA-N 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000007431 microscopic evaluation Methods 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 230000010355 oscillation Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10H—ELECTROPHONIC MUSICAL INSTRUMENTS; INSTRUMENTS IN WHICH THE TONES ARE GENERATED BY ELECTROMECHANICAL MEANS OR ELECTRONIC GENERATORS, OR IN WHICH THE TONES ARE SYNTHESISED FROM A DATA STORE
- G10H1/00—Details of electrophonic musical instruments
- G10H1/32—Constructional details
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/48—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use
- G10L25/51—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10H—ELECTROPHONIC MUSICAL INSTRUMENTS; INSTRUMENTS IN WHICH THE TONES ARE GENERATED BY ELECTROMECHANICAL MEANS OR ELECTRONIC GENERATORS, OR IN WHICH THE TONES ARE SYNTHESISED FROM A DATA STORE
- G10H2210/00—Aspects or methods of musical processing having intrinsic musical character, i.e. involving musical theory or musical parameters or relying on musical knowledge, as applied in electrophonic musical tools or instruments
- G10H2210/031—Musical analysis, i.e. isolation, extraction or identification of musical elements or musical parameters from a raw acoustic signal or from an encoded audio signal
- G10H2210/091—Musical analysis, i.e. isolation, extraction or identification of musical elements or musical parameters from a raw acoustic signal or from an encoded audio signal for performance evaluation, i.e. judging, grading or scoring the musical qualities or faithfulness of a performance, e.g. with respect to pitch, tempo or other timings of a reference performance
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10H—ELECTROPHONIC MUSICAL INSTRUMENTS; INSTRUMENTS IN WHICH THE TONES ARE GENERATED BY ELECTROMECHANICAL MEANS OR ELECTRONIC GENERATORS, OR IN WHICH THE TONES ARE SYNTHESISED FROM A DATA STORE
- G10H2220/00—Input/output interfacing specifically adapted for electrophonic musical tools or instruments
- G10H2220/155—User input interfaces for electrophonic musical instruments
- G10H2220/211—User input interfaces for electrophonic musical instruments for microphones, i.e. control of musical parameters either directly from microphone signals or by physically associated peripherals, e.g. karaoke control switches or rhythm sensing accelerometer within the microphone casing
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10H—ELECTROPHONIC MUSICAL INSTRUMENTS; INSTRUMENTS IN WHICH THE TONES ARE GENERATED BY ELECTROMECHANICAL MEANS OR ELECTRONIC GENERATORS, OR IN WHICH THE TONES ARE SYNTHESISED FROM A DATA STORE
- G10H2240/00—Data organisation or data communication aspects, specifically adapted for electrophonic musical tools or instruments
- G10H2240/075—Musical metadata derived from musical analysis or for use in electrophonic musical instruments
- G10H2240/081—Genre classification, i.e. descriptive metadata for classification or selection of musical pieces according to style
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10H—ELECTROPHONIC MUSICAL INSTRUMENTS; INSTRUMENTS IN WHICH THE TONES ARE GENERATED BY ELECTROMECHANICAL MEANS OR ELECTRONIC GENERATORS, OR IN WHICH THE TONES ARE SYNTHESISED FROM A DATA STORE
- G10H2250/00—Aspects of algorithms or signal processing methods without intrinsic musical character, yet specifically adapted for or used in electrophonic musical processing
- G10H2250/311—Neural networks for electrophonic musical instruments or musical processing, e.g. for musical recognition or control, automatic composition or improvisation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Computational Linguistics (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Electrophonic Musical Instruments (AREA)
Abstract
The invention provides a piano playing tone quality evaluation system and method combined with music. The system comprises a piano music library, a microphone array recording device, a high-quality microphone recording device, a music database, an expert listening evaluation module, a signal feature extraction module, a sample library, a signal feature analysis module and a tone quality evaluation module. After the system is established, a user only needs to select a song with preference to the music, and the user plays the song on the piano to be evaluated, so that the system can intelligently process collected playing sound signals and output an evaluation score as an evaluation result of the piano playing sound quality. The system provided by the invention combines the curved wind factors, analyzes various signal characteristics such as airspace, time domain, frequency domain, time-frequency diagram and the like, and intelligently outputs the evaluation result in the use process.
Description
Technical Field
The invention relates to the cross technical field of music acoustics, psychoacoustics, signal processing, computer mode identification and the like, in particular to a piano playing tone quality evaluation system and method combining music.
Background
With the continuous development of economy and society, the living standard of people is rapidly improved, the spiritual culture is concerned and paid more and more attention, and the demand is increasingly increased. Among them, the music art is the most contacted aspect in people's daily life, and the piano as the "king of musical instrument" occupies a large part of the proportion of playing the music art. Because the piano has a complex structure and a wide range of sound, the difficulty of researching and evaluating the tone quality and tone color is high. At present, the quality of the sound quality of the piano basically depends on subjective judgment, namely, experts with certain piano music theory knowledge and playing experience listen and score according to some indexes. The evaluation method can cause the problems of fluctuation of evaluation accuracy and the like due to auditory fatigue of testers, personal preference and different on-site listening environments. Furthermore, in many real-time situations, the labor and time costs of listening tests are also relatively expensive.
With the development of computer science and the research on the combination of signal processing technology and music acoustics, people begin to analyze the characteristics of various aspects of musical instruments from a rational and scientific point of view. The music signal characteristics of musical instrument playing sound signals can be accurately extracted by applying digital signal processing technologies such as wavelet analysis and the like or machine learning knowledge such as neural networks, and most of the previous researches are applied to aspects such as musical instrument identification, playing correct and incorrect evaluation and the like by using the characteristics or simply evaluating the sound quality by directly analyzing the characteristics, so that the sound quality evaluation cannot be automatically and intelligently carried out. If the subjective priori knowledge and the objective signal characteristic analysis can be combined to find the mapping relation between the subjective priori knowledge and the objective signal characteristic analysis, then an intelligent evaluation system is designed by utilizing the technologies of neural network, fuzzy reasoning and the like which are closer to the human cognition and judgment mechanism, finally, only a section of piano playing sound signal is input to a computer, the labor cost can be saved, and the piano playing sound quality does not need to be evaluated manually.
In the aspect of music signal feature extraction and analysis, more and more attention and research are paid to the relationship between the subjective priori knowledge of piano music and the objective digital signal features. In the previous researches, most of the researches are based on the music recognition of contents, and most of the researches are to collect signals of piano playing sounds by using a single microphone, and in the researches combined with a microphone array, most of the researches are to collect signals of a spatial field, apply the signals to sound field simulation and sound field reproduction of the spatial field, rarely use the microphone array to receive signals of the piano playing sounds at different positions, namely measurement and collection of spatial signals, wherein the spatial signals are related to overall auditory perception information, and meanwhile, in the researches on the influence of sound board characteristics on the sound quality of the piano, the spatial characteristics of the spatial signals are closely related to the sound board characteristics. Therefore, considering the collection of spatial domain signals of a plurality of positions and the comprehensive analysis, the tone quality of one piano can be more comprehensively evaluated.
On the other hand, in the case of the piano tone quality study, only a single note, a double note, or a simple performance sample without a melody label is often used in the previous study in analyzing music data, and the subjective preference of the player, that is, a factor of the melody is not considered. In the social environment, the individual requirements are more emphasized by piano manufacturers and sellers, and each piano fan has own preference, so that the piano most suitable for own habits and preferences is expected to be found. And whether the beginners or the experienced experts, when players play piano music of different styles on the same piano, the tone quality of the piano is different. Therefore, the quality of a piano is evaluated, and a factor of personalized characteristics such as a music style should be considered.
Disclosure of Invention
The invention aims to overcome the defects that the analysis is not comprehensive enough due to the fact that the piano tone quality evaluation depends on subjective evaluation and the music is not considered, and solves the problem that the spatial domain characteristic analysis of piano playing tone signals is lacked due to the fact that only a single microphone is used for data acquisition, and provides a piano playing tone quality evaluation system combined with the music and a playing tone signal acquisition device combining a microphone array recording device and a high-quality microphone recording device. Meanwhile, the tone quality evaluation module is constructed by adopting the fuzzy neural network, after a playing sound signal of a tune with a specific style played on one piano by a user is received, the corresponding characteristics of the data in a domain, a time domain, a frequency domain and a time-frequency diagram are analyzed, the obtained characteristic vector and the tune label are used as the input of the fuzzy neural network, a score is finally output, and the tone quality evaluation is carried out on the played piano when the tune is selected.
In order to realize the purpose and the function, the system provided by the invention needs to be subjected to the process of system establishment before use, namely, collecting playing sound signals of different music styles on different pianos, collecting expert subjective evaluation data, extracting and analyzing signal characteristics, determining the structure of a fuzzy neural network and training; the trained fuzzy neural network can be used. The specific technical scheme of the invention is as follows.
A piano playing tone quality evaluation system combined with music comprises a piano music library, a microphone array recording device, a high-quality microphone recording device, a music database, an expert listening evaluation module, a signal characteristic extraction module, a sample library, a signal characteristic analysis module and a tone quality evaluation module;
the piano music library is provided with music labels, and the music labels refer to the serial number marking of each music on the basis of music classification; the piano music library has two functions, namely, in the system establishing process, a large number of piano music resources are provided for subsequent analysis and training of a fuzzy neural network; secondly, in the using process of the user, selectable piano music is provided for the user;
the microphone array recording device comprises microphones arranged at different spatial positions, and a plurality of microphones collect playing sound signals, namely space-domain signals, of the piano at different positions so as to realize macroscopically analyzing the playing sound quality of the piano;
the signal characteristic extraction module extracts macroscopic and microscopic signal characteristics through signal characteristic extraction of audio files formed by piano playing sounds collected by the microphone array recording device and the high-quality microphone recording device, wherein the signal characteristics comprise space domain characteristics, time domain characteristics, frequency domain characteristics and time-frequency diagram characteristics; then establishing a sample library for storing input samples of the fuzzy neural network, wherein the input samples comprise two sample types of training samples and samples to be evaluated; the sample content is a signal feature vector and comprises the extracted signal features and corresponding curved wind labels;
the signal characteristic analysis module realizes the function of establishing a fuzzy set and a fuzzy inference rule, and comprises the steps of carrying out statistical comparison analysis on the extracted signal characteristics and subjective evaluation data obtained from the expert listening evaluation module, thereby establishing a fuzzy set of the signal characteristics and the subjective evaluation data, and simultaneously establishing a fuzzy inference rule of an evaluation process;
the tone quality evaluation module realizes the function of outputting tone quality evaluation scores after the samples in the sample library are processed by the fuzzy neural network; before the tone quality evaluation module is used, the structure of a fuzzy neural network is determined by adopting a fuzzy inference rule established by the signal characteristic analysis module, all samples of the sample library are used as training samples of the fuzzy neural network, subjective evaluation data of an audio file corresponding to each sample are used as expected output, namely a supervision signal, and then the fuzzy neural network is trained; after the network training is finished, the tone quality evaluation module can be used; when the user uses the tone quality evaluation module, the function of intelligently outputting the evaluation score after obtaining the sample to be evaluated can be realized without manually obtaining subjective evaluation data.
Based on the technical scheme, the evaluation method of the piano playing sound quality evaluation system combining the music comprises two processes, namely a system establishing process and a user using process.
Before the system is used, playing sound signals of different music on different pianos need to be collected, expert subjective evaluation data are collected, signal characteristics are extracted and analyzed, and the structure of a fuzzy neural network is determined and trained.
The steps of the system setup procedure are as follows:
(1) the piano music library with the music labels is established by analyzing the representative music classification rules, and the music labels are used for numbering and marking the types of the music on the basis of music classification.
(2) The performance sound signals played on a plurality of pianos are recorded on site using a microphone array recording apparatus and a high-quality microphone recording apparatus, and the played tunes include all the tunes in a piano tune library. The performance sound signals collected by the microphone array recording device are processed to form a multi-channel audio file, and the performance sound signals collected by the high-quality microphone recording device are correspondingly processed to form a high-fidelity audio file.
(3) The high-fidelity audio file is played back in a listening room, a plurality of professionals are allowed to carry out listening experiments, the quality of the sound quality of the professional is evaluated, subjective evaluation data are collected, and the subjective evaluation data, the audio file and the music note tag which are obtained through statistics are correspondingly stored in a music database.
(4) And based on the short-time stationarity of the audio signal, carrying out pre-emphasis processing and framing processing on the multi-channel audio file and the high-fidelity audio file.
(5) Inputting all audio files in the music database into a signal feature extraction module to extract signal features, establishing a sample library, forming signal feature vectors by the extracted signal features and corresponding music labels, and correspondingly storing the signal feature vectors into the sample library to be used as training samples.
(6) Counting the extracted signal characteristics, and carrying out comparative analysis on the signal characteristics and the subjective evaluation data in the step (3) so as to establish a fuzzy set of the signal characteristics and the subjective evaluation data and establish a fuzzy inference rule of an evaluation process; the fuzzy set and fuzzy inference rules refer to conditions required for judging the sound quality and the description of judgment logic in fuzzy mathematics; as can be described as follows: if the fundamental frequency harmonic proportion is in the range of x1 and the spatial balance is in the range of y1, the sound quality evaluation score is in the range of z 1; x1, y1, z1 all belong to respective fuzzy sets X, Y, Z.
(7) And (4) determining the structure of the fuzzy neural network according to the size of the signal feature vector obtained in the step (5) and the fuzzy inference rule obtained by analysis in the step (6), taking all samples of the sample library as training samples of the fuzzy neural network, taking subjective evaluation data of the audio file corresponding to each sample as expected output, namely a supervision signal, then training the fuzzy neural network, and finishing the establishment of the tone quality evaluation module.
After the system is built, a user can use the system, and the function of intelligently outputting the evaluation score after the sample to be evaluated is obtained can be realized without manually obtaining subjective evaluation data. The steps of the user using the system are as follows:
(1) let the user select a tune in the piano tune library and play on the piano to be evaluated, the microphone array recording apparatus and the high-quality microphone recording apparatus simultaneously collect the performance tone signals.
(2) Processing the playing sound signals to form audio files, and storing the audio files and the music labels into a music database;
(3) the audio file is input into a signal feature extraction module after being preprocessed, the extracted signal features and the music wind labels form signal feature vectors together, and the signal feature vectors are stored in a sample library to serve as samples to be evaluated.
(4) The sample to be evaluated is input into the established sound quality evaluation module, and finally the sound quality evaluation module outputs an evaluation score in the range of [0,100] as the performance sound quality evaluation result of the selected piano.
Compared with the existing tone quality evaluation system, the invention has the following advantages:
(1) according to the piano playing tone quality evaluation system combined with the music, after the system is established, only a section of piano playing tone signal needs to be input into the computer, the piano playing tone quality can be evaluated without subsequent manual work, and compared with the manual subjective evaluation of experts, the piano playing tone quality evaluation system combined with the music greatly improves convenience and intelligence and reduces labor cost and time cost.
(2) The system provided by the invention fully considers subjective evaluation data and objective signal characteristics in the system establishing process, wherein the objective signal characteristics comprise space domain characteristics, time domain characteristics, frequency domain characteristics and time-frequency diagram characteristics, and the multi-characteristic analysis method greatly improves the objectivity of the evaluation system and avoids the problems of evaluation standard floating and the like caused by artificial subjective evaluation.
(3) The system provided by the invention also considers the factors of the music which have personalized characteristics, and brings the music label into the piano for playing, and after the system processing, the user can obtain the sound quality of the music which is preferred by the user on the piano to be evaluated. Therefore, the user can better assist the user in selecting the piano really suitable for the user by comparing the evaluation scores of the plurality of intended pianos on the favorite music, so that the playing effect is optimal; and according to the preference of a user, a piano manufacturer can use the system to assist in debugging various parameters of the piano when producing the personalized custom piano.
Drawings
Fig. 1 is a block diagram showing the overall configuration of a piano playing sound quality evaluation system incorporating music.
Fig. 2 is a block diagram of a microphone array recording apparatus.
FIG. 3 is a flow diagram of an example system set up.
FIG. 4 is an example user usage flow diagram.
Detailed Description
In order to make the objects, technical solutions, innovative points and advantages of the present invention more apparent, embodiments of the present invention are further described below with reference to the accompanying drawings.
The following further describes embodiments of the present invention with reference to the drawings, but the practice of the present invention is not limited thereto.
As shown in fig. 1, the system of the present invention comprises nine modules: the system comprises a piano music library, a microphone array recording device, a high-quality microphone recording device, a music database, an expert listening evaluation module, a signal characteristic extraction module, a sample library, a signal characteristic analysis module and a tone quality evaluation module.
The piano performance sound quality evaluation system combined with the music wind mainly comprises the following parts:
(1) piano music library: the piano music book comprises a plurality of piano music books with music book labels, wherein the music book labels are used for numbering and marking each music book on the basis of music book classification.
(2) Microphone array recording device: the device is used for collecting spatial signals of piano playing tones for macroscopic analysis.
(3) High-quality microphone recording device: the signal is used for acquiring the piano high-fidelity performance sound for microscopic analysis.
(4) An expert listening evaluation module: in the system establishing process, playing sound signals of all piano songs on different pianos in a piano song library are required to be collected, and then the playing sound signals are correspondingly processed to form audio files; then, subjective evaluation data is obtained through the process of audio file playback and expert listening evaluation.
(5) A music database: correspondingly storing the music style label, the audio file and the subjective evaluation data into a music database; the process of playback of audio files and expert listening evaluation is not needed during the use of the system by the user, so that the data stored in the music library only comprises the audio files and the music labels.
(6) A signal feature extraction module: after the audio files of the playing sound in the music database are preprocessed, the macroscopic and microscopic signal characteristics of the piano playing sound are extracted through the module, and the signal characteristics mainly comprise space domain characteristics, time domain characteristics, frequency domain characteristics and time-frequency diagram characteristics.
(7) Sample library: the part is used for storing input samples of the fuzzy neural network, the sample content is a signal feature vector and comprises extracted signal features and corresponding curved wind labels, and the sample types comprise training samples and samples to be evaluated.
(8) A signal characteristic analysis module: the module is mainly used for realizing the function of establishing the fuzzy inference rule in the system establishing process, and comprises the steps of carrying out statistic comparison analysis on the extracted signal characteristics and subjective evaluation data, establishing a fuzzy set of the signal characteristics and the subjective evaluation data and establishing the fuzzy inference rule in the evaluation process.
(9) And a tone quality evaluation module: the module realizes the function of outputting the tone quality evaluation score after the samples in the sample library are processed by the fuzzy neural network. Before the module is used, the fuzzy inference rule established by the signal characteristic analysis module in the step (8) is adopted to determine the structure of the fuzzy neural network, all samples in the sample library in the step (7) are used as training samples of the fuzzy neural network, subjective evaluation data of an audio file corresponding to each sample is used as expected output, namely a supervision signal, and then the fuzzy neural network is trained. After the network is trained, the tone quality evaluation module can be used. When the module is used by a user, the function of intelligently outputting the evaluation score after the sample to be evaluated is obtained can be realized without manually obtaining subjective evaluation data.
Before the system provided by the invention is used, the process of system establishment is required, namely, playing sound signals of different music on different pianos are collected, expert subjective evaluation data are collected, signal characteristics are extracted and analyzed, and the structure of a fuzzy neural network is determined and trained; the trained fuzzy neural network can be used. Thus, the system set-up process includes all nine modules, while the user usage process includes only seven of the modules: the system comprises a piano music library, a microphone array recording device, a high-quality microphone recording device, a music database, a signal characteristic extraction module, a sample library and a tone quality evaluation module.
The piano tune library in this example may include four types of styles formed by classifying the piano tune development periods: the song label is respectively numbered 1,2,3 and 4 in the baroque style, the classical style, the romantic style and the modern style; selecting three typical exercise songs from each type of style; thus, a total of 12 piano tunes were included in the tune library.
The microphone array recording device in this embodiment is shown in fig. 2, and includes 18 microphones, an interface assembly board, a USB to serial port signal transmission control circuit, an amplification band-pass circuit, an a/D module, and an ARM signal processing and storing module. The amplifying band-pass circuit, the A/D module and the ARM signal processing and storing module are integrated into a set circuit, each set circuit board receives playing sound signals collected by 3 microphones, and 6 set circuit boards in total comprise 1 host set circuit and 5 slave set circuits; the sampling frequency is 100kHz, and the quantization precision is 12 bits; the USB-to-serial port signal transmission control circuit is mainly responsible for receiving signals of starting and stopping recording at the computer end, when the computer sends a command for starting receiving, the command is transmitted to the host through the module, the host simultaneously transmits the command to each slave through a signal wire connected with the slave, and then other modules start working; data in the ARM signal processing storage module are stored on the SD storage card in a bin file mode, and are converted into multichannel wav audio files through corresponding programs.
The high-quality microphone recording device comprises a single microphone capable of collecting high-fidelity playing sound signals, and is combined with Adobe Audio software to generate a high-fidelity wav audio file, the selected sampling frequency is 96kHz, and the quantization precision is 16 bits.
The music database in the embodiment is mainly used for storing multi-channel wav audio files and high-fidelity wav audio files, the audio files, subjective evaluation data obtained after the audio files are subjected to the expert listening evaluation module, and music labels are correspondingly stored in the database, so that the music database with various styles, multiple pianos and different sound quality evaluation score labels is established.
And wherein the expert listening evaluation module mainly comprises the following components: the high-fidelity wav audio file interception part is played back in a listening room, 5 experts are invited to perform listening experiments, each intercepted audio is evaluated according to indexes such as sound stability, richness, brightness, fullness and the like, then tone quality between every two audio files is compared and scored integrally, and the collected subjective evaluation data is counted and analyzed.
The music feature extraction module in the embodiment mainly realizes the function of extracting and obtaining features in the aspects of airspace, time domain, frequency domain, time-frequency diagram and the like by taking an audio file in a music database as input; the relationship between these features and the psychoacoustic index is as follows:
1) macroscopic spatial domain signal characteristics such as spatial balance extracted based on the multi-channel wav audio file are related to the stability of sound and the naturalness of a sound area and a transition area;
2) microscopic time domain envelope characteristics such as the oscillation starting time and the single tone time value extracted based on the high-fidelity wav audio file are related to the tone brightness or low tone quality of the sound, and the tone quality is directly influenced;
3) microscopic frequency domain characteristics such as amplitude and amplitude proportion of fundamental frequency and overtone extracted based on the high-fidelity wav audio file are related to timbre expressive force of sound;
4) the time-frequency diagram shows that the overall characteristics of fundamental frequency and overtone are related to the tone fullness and harmony of sound.
The sample library in the embodiment is used for storing input samples of the tone quality evaluation module, and comprises two sample types, namely a training sample and a sample to be evaluated; the sample content is a signal feature vector comprising the extracted signal features and corresponding melody tags.
The tone quality evaluation module in the embodiment is a main module of the system, the main structure of the tone quality evaluation module is a multi-input single-output five-layer neural network, and the selection of internal logic and weight functions depends on a fuzzy set and a fuzzy inference rule established by a signal characteristic analysis module.
The music signal characteristic analysis module mainly realizes the function of establishing a fuzzy inference rule, and comprises the steps of carrying out statistic comparison analysis on the extracted signal characteristics and subjective evaluation data, thereby establishing a fuzzy set of the signal characteristics and the subjective evaluation data, and simultaneously establishing the fuzzy inference rule of an evaluation process; the fuzzy set and the fuzzy rule refer to conditions required for judging the sound quality and the description of a judgment logic in fuzzy mathematics; as can be described as follows: if the fundamental frequency harmonic proportion is in the range of x1 and the spatial balance is in the range of y1, the sound quality evaluation score is in the range of z 1. x1, y1, z1 all belong to respective fuzzy sets X, Y, Z.
Before the system is used, the system needs to be subjected to the process of system establishment, namely, playing sound signals of different music styles on different pianos are collected, expert subjective evaluation data are collected, signal characteristics are extracted and analyzed, and the structure of a fuzzy neural network is determined and trained; the trained fuzzy neural network can be used. A flow chart of the system set-up procedure is shown in figure 3.
The system establishment process is as follows:
(1) establishing a piano music library: as described above, a total of 12 piano songs of four genres are included in the library.
(2) The recording device is turned on, and the player plays the music in the music library: the performance sound signals played on a plurality of pianos are recorded on site using a microphone array recording apparatus and a high-quality microphone recording apparatus, and the played tunes include all the tunes in a piano tune library.
(3) Forming an audio file: the performance sound signals collected by the microphone array recording device are processed to form a multi-channel audio file, and the performance sound signals collected by the high-quality microphone recording device are correspondingly processed to form a high-fidelity audio file. In order to correspond to each exercise music of each style in the piano music library of this example, thus letting 3 players play on 4 distinctly different pianos, 48 hi-fi audio files and 48 16-channel audio files are obtained in total.
(4) Audio file playback, expert listening evaluation, music database establishment: the high-fidelity audio file is played back in a listening room, a plurality of professionals are allowed to carry out listening experiments, the quality of the sound quality of the professional is evaluated, subjective evaluation data are collected, and the subjective evaluation data, the audio file and the music note tag which are obtained through statistics are correspondingly stored in a music database.
(5) Preprocessing an audio file: and based on the short-time stationarity of the audio signal, carrying out pre-emphasis processing and framing processing on the multi-channel audio file and the high-fidelity audio file.
(6) Extracting signal features to form training samples: inputting all audio files in the music database into a signal feature extraction module to extract signal features, establishing a sample library, forming signal feature vectors by the extracted signal features and corresponding music labels, and correspondingly storing the signal feature vectors into the sample library to be used as training samples.
(7) Analyzing the signal characteristics and subjective evaluation data, and establishing a fuzzy set and a fuzzy inference rule: counting the extracted signal characteristics, and carrying out comparative analysis on the extracted signal characteristics and the subjective evaluation data obtained in the step (4), so as to establish a fuzzy set of the signal characteristics and the subjective evaluation data and establish a fuzzy inference rule of an evaluation process; the fuzzy set and fuzzy inference rules refer to conditions required for judging the sound quality and the description of judgment logic in fuzzy mathematics; as can be described as follows: if the fundamental frequency harmonic proportion is in the range of x1 and the spatial balance is in the range of y1, the sound quality evaluation score is in the range of z 1; x1, y1, z1 all belong to respective fuzzy sets X, Y, Z.
(8) Determining the structure of the fuzzy neural network, training the fuzzy neural network: and (4) determining the structure of the fuzzy neural network according to the size of the signal feature vector obtained in the step (6) and the fuzzy inference rule obtained by analysis in the step (7), taking all samples of the sample library as training samples of the fuzzy neural network, taking subjective evaluation data of the audio file corresponding to each sample as expected output, namely a supervision signal, and then training the fuzzy neural network.
And after the verification and verification evaluation result is verified to be optimal, the tone quality evaluation module is established. After the system is successfully established, a user can use the system, and the function of intelligently outputting evaluation scores after obtaining samples to be evaluated can be realized without manually obtaining subjective evaluation data; a flow chart of the user using the system is shown in fig. 4.
The flow of the system using process of the user is as follows:
(1) user selection of preferred music: the user selects a preferred tune in the piano tune library, for example, a romantic style tune, the tune label number is set to 3, and the selected tune is correctly played on the piano desired to obtain the sound quality evaluation.
(2) The recording device is turned on, and the user plays the tune of the selected tune on the piano to be evaluated: the microphone array recording device and the high-quality microphone recording device simultaneously acquire performance sound signals of a user.
(3) Forming an audio file, and storing the audio file in a music database by combining with a music label: and after the performance is finished, processing the collected performance sound signals to form audio files, and storing the audio files and the music labels into a music database.
(4) Preprocessing an audio file: and performing pre-emphasis processing and framing processing on the audio file based on the short-time stationarity of the audio signal.
(5) Extracting signal characteristics to form a sample to be evaluated: and inputting the preprocessed audio file into a signal feature extraction module, forming a signal feature vector by the extracted signal feature and the song label, and storing the signal feature vector into a sample library to be used as a sample to be evaluated.
(6) And (3) performing sound quality evaluation, and outputting a sound quality evaluation score: inputting the sample to be evaluated into the established sound quality evaluation module, and finally outputting an evaluation score, such as 65 scores, within the range of [0,100], by the sound quality evaluation module, wherein the evaluation score is the performance sound quality evaluation result of the played piano under the selected curved wind.
Therefore, the user can compare the evaluation scores of multiple intended pianos on the favorite music so as to assist in selecting the piano really suitable for the user and enable the playing to achieve the best effect; according to the preference of users, piano manufacturers can use the system to assist in debugging various parameters of the piano when producing personalized custom pianos, so that a large part of manual evaluation cost and time cost are reduced.
The above embodiments are intended to be preferred embodiments of the present invention, but the present invention is not limited to the above embodiments, and any other changes, modifications, substitutions, combinations, and simplifications which do not depart from the spirit and principle of the present invention should be construed as equivalents thereof, and all such changes, modifications, substitutions, combinations, and simplifications are intended to be included in the scope of the present invention.
Claims (2)
1. A piano playing tone quality evaluation system combined with music is characterized by comprising a piano music library, a microphone array recording device, a high-quality microphone recording device, a music database, an expert listening evaluation module, a signal feature extraction module, a sample library, a signal feature analysis module and a tone quality evaluation module;
the piano music library is provided with music labels, and the music labels refer to the serial number marking of each music on the basis of music classification; the piano music library has two functions, namely, in the system establishing process, a large number of piano music resources are provided for subsequent analysis and training of a fuzzy neural network; secondly, in the using process of the user, selectable piano music is provided for the user;
the microphone array recording device comprises microphones arranged at different spatial positions, and a plurality of microphones collect playing sound signals, namely space-domain signals, of the piano at different positions so as to realize macroscopically analyzing the playing sound quality of the piano;
the microphone array recording device is used for acquiring spatial signals of piano playing tones for macroscopic analysis; the high-quality microphone recording device comprises: the signal used for collecting piano high fidelity performance sound is used for analyzing on a microcosmic scale; the high-quality microphone recording device comprises a single microphone capable of collecting high-fidelity playing sound signals, and is combined with Adobe Audio software to generate a high-fidelity wav audio file;
the signal characteristic extraction module extracts macroscopic and microscopic signal characteristics through signal characteristic extraction of audio files formed by piano playing sounds collected by the microphone array recording device and the high-quality microphone recording device, wherein the signal characteristics comprise space domain characteristics, time domain characteristics, frequency domain characteristics and time-frequency diagram characteristics; then establishing a sample library for storing input samples of the fuzzy neural network, wherein the input samples comprise two sample types of training samples and samples to be evaluated; the sample content is a signal feature vector and comprises the extracted signal features and corresponding curved wind labels;
the signal characteristic analysis module realizes the function of establishing a fuzzy set and a fuzzy inference rule, and comprises the steps of carrying out statistical comparison analysis on the extracted signal characteristics and subjective evaluation data obtained from the expert listening evaluation module, thereby establishing a fuzzy set of the signal characteristics and the subjective evaluation data, and simultaneously establishing a fuzzy inference rule of an evaluation process;
the tone quality evaluation module realizes the function of outputting tone quality evaluation scores after the samples in the sample library are processed by the fuzzy neural network; before the tone quality evaluation module is used, the structure of a fuzzy neural network is determined by adopting a fuzzy inference rule established by the signal characteristic analysis module, all samples of the sample library are used as training samples of the fuzzy neural network, subjective evaluation data of an audio file corresponding to each sample are used as expected output, namely a supervision signal, and then the fuzzy neural network is trained; after the network training is finished, the tone quality evaluation module can be used; when the user uses the tone quality evaluation module, the function of intelligently outputting the evaluation score after obtaining the sample to be evaluated can be realized without manually obtaining subjective evaluation data.
2. An evaluation method of piano playing sound quality evaluation system using the combination of the music as claimed in claim 1, characterized by comprising a system establishing process and a user using process; before the system is used, playing sound signals of different music styles on different pianos need to be collected, expert subjective evaluation data are collected, signal characteristics are extracted and analyzed, and the structure of a fuzzy neural network is determined and trained;
the steps of the system setup procedure are as follows:
(1) establishing a piano music library with music labels by analyzing representative music classification rules, wherein the music labels refer to numbering and marking of each music on the basis of music classification;
(2) recording playing sound signals played on a plurality of pianos on site by using a microphone array recording device and a high-quality microphone recording device, wherein the played music comprises all music in a piano music library; processing the performance sound signals collected by the microphone array recording device to form a multi-channel audio file, and correspondingly processing the performance sound signals collected by the high-quality microphone recording device to form a high-fidelity audio file;
(3) the method comprises the steps that a high-fidelity audio file is played back in a listening room, a plurality of professionals carry out listening experiments, the quality of the sound quality of the professional is evaluated, subjective evaluation data are collected, and the subjective evaluation data, the audio file and a music note tag which are obtained through statistics are correspondingly stored in a music database;
(4) based on the short-time stationarity of the audio signal, carrying out pre-emphasis processing and framing processing on a multi-channel audio file and a high-fidelity audio file;
(5) inputting all audio files in a music database into a signal feature extraction module to extract signal features, establishing a sample library, forming signal feature vectors by the extracted signal features and corresponding music labels, and correspondingly storing the signal feature vectors into the sample library to serve as training samples;
(6) counting the extracted signal characteristics, and carrying out comparative analysis on the signal characteristics and the subjective evaluation data in the step (3) so as to establish a fuzzy set of the signal characteristics and the subjective evaluation data and establish a fuzzy inference rule of an evaluation process; the fuzzy set and fuzzy inference rules refer to conditions required for judging the sound quality and the description of judgment logic in fuzzy mathematics; as can be described as follows: if the fundamental frequency harmonic proportion is in the range of x1 and the spatial balance is in the range of y1, the sound quality evaluation score is in the range of z 1; x1, y1, z1 all belong to respective fuzzy sets X, Y, Z;
(7) determining the structure of the fuzzy neural network according to the size of the signal feature vector obtained in the step (5) and the fuzzy inference rule obtained by analysis in the step (6), taking all samples of a sample library as training samples of the fuzzy neural network, taking subjective evaluation data of an audio file corresponding to each sample as expected output, namely a supervision signal, then training the fuzzy neural network, and finishing the establishment of a tone quality evaluation module;
after the system is built, a user can use the system, and the function of intelligently outputting evaluation scores after a sample to be evaluated is obtained can be realized without manually obtaining subjective evaluation data;
the steps of the user using the system are as follows:
(1) enabling a user to select a song in a piano song library and play on a piano to be evaluated, and simultaneously acquiring playing sound signals by the microphone array recording device and the high-quality microphone recording device;
(2) processing the playing sound signals to form audio files, and storing the audio files and the music labels into a music database;
(3) preprocessing an audio file, inputting the preprocessed audio file into a signal feature extraction module, forming a signal feature vector by the extracted signal feature and a music wind label, and storing the signal feature vector in a sample library as a sample to be evaluated;
(4) the sample to be evaluated is input into the established sound quality evaluation module, and finally the sound quality evaluation module outputs an evaluation score in the range of [0,100] as the performance sound quality evaluation result of the selected piano.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810311753.2A CN108735192B (en) | 2018-04-09 | 2018-04-09 | System and method for evaluating piano playing tone quality by combining music |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810311753.2A CN108735192B (en) | 2018-04-09 | 2018-04-09 | System and method for evaluating piano playing tone quality by combining music |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108735192A CN108735192A (en) | 2018-11-02 |
| CN108735192B true CN108735192B (en) | 2021-09-21 |
Family
ID=63941242
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810311753.2A Expired - Fee Related CN108735192B (en) | 2018-04-09 | 2018-04-09 | System and method for evaluating piano playing tone quality by combining music |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108735192B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109991842B (en) * | 2019-03-14 | 2022-07-01 | 合肥工业大学 | Piano tuning method and system based on neural network |
| CN110516104A (en) * | 2019-08-27 | 2019-11-29 | 腾讯音乐娱乐科技（深圳）有限公司 | Song recommendations method, apparatus and computer storage medium |
| CN110675879B (en) * | 2019-09-04 | 2023-06-23 | 平安科技（深圳）有限公司 | Audio evaluation method, system, equipment and storage medium based on big data |
| CN113108894A (en) * | 2021-03-30 | 2021-07-13 | 华南理工大学 | Transient sound field measuring system and method of musical instrument soundboard based on audio array acquisition |
| CN114842876A (en) * | 2022-04-27 | 2022-08-02 | 厦门大学 | Lute tone quality evaluation method based on vibroacoustics |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105719661A (en) * | 2016-01-29 | 2016-06-29 | 西安交通大学 | Automatic discrimination method for playing timbre of string instrument |
| CN107146497A (en) * | 2016-08-02 | 2017-09-08 | 浙江大学 | A kind of piano test for confirmation of the grade points-scoring system |
-
2018
- 2018-04-09 CN CN201810311753.2A patent/CN108735192B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105719661A (en) * | 2016-01-29 | 2016-06-29 | 西安交通大学 | Automatic discrimination method for playing timbre of string instrument |
| CN107146497A (en) * | 2016-08-02 | 2017-09-08 | 浙江大学 | A kind of piano test for confirmation of the grade points-scoring system |
Non-Patent Citations (4)
| Title |
|---|
| "musical instrument recognition in user-generated videos using a multimodal convolutional neural network architecture";Olga Slizovskaia et al.;《ACM on international conference on Multimedia retrieval》;20170609;全文 * |
| "声乐演唱中音色的评价标准";黄阿罗等;《黄河之声》;20171130(第11期);全文 * |
| "钢琴演奏评价***研究";刘乐;《中国优秀博硕士学位论文全文数据库（硕士） 信息科技辑》;20060815;全文 * |
| "钢琴音乐信号的特征识别";吴晶晶;《中国优秀硕士学位论文全文数据库 哲学与人文科学辑》;20100715;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108735192A (en) | 2018-11-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108735192B (en) | System and method for evaluating piano playing tone quality by combining music | |
| CN110019931B (en) | Audio classification method and device, intelligent equipment and storage medium | |
| CN102610222B (en) | Music transcription method, system and device | |
| De Poli et al. | Sonological models for timbre characterization | |
| CN109147804A (en) | A kind of acoustic feature processing method and system based on deep learning | |
| US9875304B2 (en) | Music selection and organization using audio fingerprints | |
| CN105719661B (en) | A kind of stringed musical instrument performance sound quality automatic distinguishing method | |
| Chau et al. | The emotional characteristics and timbre of nonsustaining instrument sounds | |
| US20150220633A1 (en) | Music selection and organization using rhythm, texture and pitch | |
| RU2013140574A (en) | SEMANTIC SOUND TRACK MIXER | |
| CN102723079B (en) | Music and chord automatic identification method based on sparse representation | |
| CN106997765B (en) | Quantitative characterization method for human voice timbre | |
| DE102012103553A1 (en) | AUDIO SYSTEM AND METHOD FOR USING ADAPTIVE INTELLIGENCE TO DISTINCT THE INFORMATION CONTENT OF AUDIOSIGNALS IN CONSUMER AUDIO AND TO CONTROL A SIGNAL PROCESSING FUNCTION | |
| Bartsch et al. | Singing voice identification using spectral envelope estimation | |
| CN109584904B (en) | Video-song audio-song name recognition modeling method applied to basic music video-song education | |
| KR101459324B1 (en) | Evaluation method of sound source and Apparatus for evaluating sound using it | |
| Chourdakis et al. | A machine-learning approach to application of intelligent artificial reverberation | |
| KR100721973B1 (en) | Method for classifying music genre using a classification algorithm | |
| CN108369800B (en) | Sound processing device | |
| GB2595222A (en) | Digital audio workstation with audio processing recommendations | |
| CN105895079A (en) | Voice data processing method and device | |
| Gounaropoulos et al. | Synthesising timbres and timbre-changes from adjectives/adverbs | |
| Setragno et al. | Feature-based characterization of violin timbre | |
| Jiang et al. | Analysis and modeling of timbre perception features of chinese musical instruments | |
| Pampalk et al. | Computational models of similarity for drum samples |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210921 |