Laboratory of Acoustic Expertise and Correction
Project: “Evaluation of the quality of musical signals perceived by microphone arrays”
Project curators:
postgraduate student Igor Kotvitsky (igorktvzk@gmail.com)
prof. Arkady Prodeus (aprodeus@gmail.com)
Active participants of the project: students of the 3rd year of study, masters of the 1st year of study
Relevance and purpose of the project. When recording music, it is preferable to record each performer separately so that later it is possible to correct the defects found by computer editing, and then combine individual records into one resulting recording.
However, this way is far from the best, because the musicians lose the “feeling of the ensemble”. In other words, it’s important for musicians to see and hear each other while performing and recording a musical composition. You can overcome this shortcoming by placing microphones near each musician.
An alternative solution that is gaining momentum is the recording of a certain point in the concert hall with the help of an acoustic antenna, which is a series of (“arrays”) microphones, united in a certain way. Although there are works [1,2,3] devoted to the study of the properties of different microphone arrays, until now it has been poorly understood which of the configurations of microphone arrays is the best in terms of the minimum distortion of musical signals. Therefore, the main goal of this project is to find the answer to this question.
Since perception of music is an individual matter, it is important for the results of the generalized results to receive as many students as possible – what is proposed to do to our students.
Project Perspectives:
This project is the first part of the design and graphic work of the disciplines “Methods of processing acoustic signals” and “Computer processing of acoustic signals” (the second part of the WGR is the project “Investigation of factors that influence the language intelligibility”).
The results of the project will be useful not only for students who are familiar with the algorithms of spatial-temporal processing of acoustic signals, but also for future consumers of such devices who learn about the potential of different types of microphone arrays.
As a result of this project students:
- get acquainted with the elements of the theory of spatial and temporal processing of signals;
- learn to organize experimental research to compare spatial-temporal processing algorithms by subjective evaluation of sound quality at the output of the corresponding software models;
- learn to correctly analyze the results obtained and formulate relevant technical reports.
Terms of Reference:
1. Setting a task
The experiments suggest listening to a series of musical signals (10 seconds saxophone solo), perceived and processed by acoustic antennas of different types in different acoustic conditions.
Two types of acoustic antennas are investigated: on the basis of sums and delays (DS) and on the basis of linearly limited minimal dispersion (LCMV) [1, 2, 3].
Acoustic conditions are characterized by different signal-to-noise ratios (-10 dB, 0 dB, and 10 dB) and different spatial arrangement of noise sources (single sources, as well as a combination of several sources).
For each signal-to-noise ratio and each type of location of noise sources, you must listen:
- reference pure signal;
- signal on the output of a single microphone;
- signal on the output of the DS antenna;
- LCMV antenna output signals (for 3 types of post-filtration).
Four situations of location of a source of noise are considered: - situations 1-3 – single noise sources located respectively at angles of 15, 45 and 90 degrees relative to the source of the signal;
- Situation 4 – a set of 4 sources, located at angles -45, +45, -120 and +120 degrees.
Students should listen to the signals through the headphones or through computer speakers (acoustic systems) and evaluate by a five-point system (points from 1 to 5): - signal quality;
- noise level;
- overall impression of the heard.
- It is allowed (and even recommended) to put fractional marks, for example, 3.7 points.
Warning! If it is unlikely that questions will be asked about the assessment of the “overall impression from the heard”, it will be somewhat more difficult to assess the “signal quality” and “noise level”. Therefore, we offer special clarifications on this subject.
One of the reasons for signal distortion (reducing its quality) is false information about the direction of the source of the signal and noise. In addition, spatial-temporal processing algorithms can be implemented insufficiently qualitatively.
As far as noise level is concerned, note that this means the noise power rating, compared with the signal strength. For listening, 10-second records are used, where the first 8 seconds sounds a mixture of music and noise, and the last 2 seconds sound “pure” noise (Fig. 1).
2. Computer programs for Matlab
Four programs are used sequentially:
- Microphone_Array_test_15.m (noise source is at an angle of 15 degrees);
- Microphone_Array_test_45.m (source of noise is at an angle of 45 degrees);
- Microphone_Array_test_90.m (noise source located at an angle of 90 degrees);
- Microphone_Array_test_istrp.m (situation with 4 sources of noise).
3. Input and output data to computer programs
Incoming data:
- cleanmusic.wav clean sound file is contained in the main folder;
- The audio files from the outputs of the simulated acoustic antennas are contained in the “10dB”, “0dB”, and “-10dB” folders.
Note: computer programs and input data are contained in the archive file:
Output data:
Output data is 4 files with names like yours_situation.mat, (for example, Ivanov_15.mat) which are stored in the folder with the main program. After the work, these 4 files must be transferred to the curators of the project (flash drive, e-mail).
Experiment duration:
The duration of the experiment with each of the 4 programs is 5-10 minutes, so the total duration of the experiment does not exceed 20-40 minutes.
4. Processing output data and passing the results to the teacher
After completing the four computer programs mentioned above, 12 charts will be obtained, as each of the four programs produces 3 graphs. These graphics should be saved to the computer’s disk and inserted into a written report.
In the report, please indicate:
- all drawings (they should be 3×4 = 12);
- your comments on the drawings (how you can interpret the results).
5. Protection of work
The results should be included in the written report and commented.
In this case, the following questions should be highlighted in the comments:
- What and why was done (that is, the relevance of the task being solved);
- Description of the organization of research – what experiences and why were asked which questions of the computer program you had to give their answers, and why these questions were raised;
Conclusions on the results of work: 1) what is done; 2) received; 3) What further needs to be done or improved?
Stages of the project:
- Start of the project (issue of the Terms of Reference) – September 3
- Transfer of results to the curators of the project – if before March 10, then the score is added 1 bonus point;
- The end of the work on this project (the first stage of the WGG) – September 17.
Project results:
As a result of the project, several articles will be written, in which the authors will list the most active students.
Literature:
- W. Zhang, P.N. Samarasinghe, H. Chen, T.D. Abhayapala. “Surround by Sound: A Review of Spatial Audio Recording and Reproduction,” Applied Sciences, No. 7, 532, 2017, pp. 1-19.
- P. Coleman, P.J. B. Jackson, and J. Francombe, “Audio object separation using microphone array beamforming,” 138th Convention, May 7, Warsaw, Poland, 2015.
- O. Schreer, G. Thomas, O.A. Niamut, J-F. Macq, A. Kochale, J-M. Batke, J. Ruiz Hidalgo, R. Oldfield, B. Shirley, G. Thallinger, “Format-agnostic Approach for Production, Delivery and Rendering of Immersive Media,” Format-agnostic Approach for Production, Delivery and Rendering of Immersive Media“, NEM Summit 2011, Torino, Italy, 27th September, 2011.