Quantifying signal purity by means of machine learning
A system includes a memory and a processor. The memory is configured to store a machine learning (ML) model. The processor is configured to (i) obtain a set of training audio signals that are labeled with respective levels of distortion, (ii) convert the training audio signals into respective images, (iii) train the ML model to estimate the levels of the distortion based on the images, (iv) receive an input audio signal, (v) convert the input audio signal into an image, and (vi) estimate a level of the distortion in the input audio signal, by applying the trained ML model to the image.
1. A system, comprising:
a memory configured to store a machine learning (ML) model; and
a processor, which is configured to:
obtain a set of training audio signals that are labeled with respective levels of distortion;
convert the training audio signals into respective images;
train the ML model to estimate the levels of the distortion based on the images;
receive an input audio signal;
convert the input audio signal into an image; and
estimate a level of the distortion in the input audio signal, by applying the trained ML model to the image.
2. The system according to claim 1 , wherein the distortion comprises a Total Harmonic Distortion (THD).
3. The system according to claim 1 , wherein the processor is configured to convert a given training audio signal into a given image by setting pixel values of the given image to represent an amplitude of the given training audio signal as a function of time.
4. The system according to claim 1 , wherein the respective images and the image are two-dimensional (2D).
5. The system according to claim 1 , wherein the respective images and the image are of three or more dimensions.
6. The system according to claim 1 , wherein the processor is configured to obtain the training audio signals by (i) receiving initial audio signals having first durations, and (ii) slicing the initial audio signals into slices having second, shorter durations, so as to produce the training audio signals.
7. The system according to claim 1 , wherein the ML model comprises a convolutional neural network (CNN).
8. The system according to claim 1 , wherein the ML model comprises a generative adversary network (GAN).
9. The system according to claim 1 , wherein the input audio signal is received from nonlinear audio processing circuitry.
10. The system according to claim 1 , wherein the ML model classifies the distortion according to the levels of distortion that label the training audio signal.
11. The system according to claim 1 , wherein the ML model estimates the level of distortion using regression.
12. The system according to claim 1 , wherein the processor is further configured to control, using the estimated level of the distortion, an audio system that produces the input audio signal.
13. A system, comprising:
a memory configured to store a machine learning (ML) model; and
a processor, which is configured to:
obtain a plurality of initial audio signals, which have first durations in a first range of durations and which are labeled with respective levels of distortion;
slice the initial audio signals into slices having second durations in a second range of durations, shorter than the first durations, so as to produce a set of training audio signals;
train the ML model to estimate the levels of the distortion based on the training audio signals, by (i) converting the training audio signals into respective images and (ii) training the ML model to estimate the levels of the distortion based on the images;
receive an input audio signal having a duration in the second range of durations; and
estimate a level of the distortion in the input audio signal by applying the trained ML model to the input audio signal.
14. The system according to claim 13 , wherein the distortion comprises a Total Harmonic Distortion (THD).
15. The system according to claim 13 , wherein the processor is configured to estimate the level of the distortion in the input audio signal by (i) converting the input audio signal into an image and (ii) applying the trained ML model to the image.
16. The system according to claim 13 , wherein the respective images are two-dimensional (2D) images.
17. The system according to claim 13 , wherein the respective images are of three or more dimensions.
18. The system according to claim 13 , wherein the ML model comprises a convolutional neural network (CNN).
19. The system according to claim 13 , wherein the ML model comprises a generative adversary network (GAN).
20. The system according to claim 13 , wherein the input audio signal is received from nonlinear audio processing circuitry.
21. The system according to claim 13 , wherein the ML model classifies the distortion according to the levels of distortion that label the training audio signal.
22. The system according to claim 13 , wherein the ML model estimates the level of distortion using regression.
23. The system according to claim 13 , wherein the processor is further configured to control, using the estimated level of the distortion, an audio system that produces the input audio signal.
24. A method, comprising:
obtaining a set of training audio signals that are labeled with respective levels of distortion;
converting the training audio signals into respective images;
training a machine learning (ML) model to estimate the levels of the distortion based on the images;
receiving an input audio signal;
converting the input audio signal into an image; and
estimating a level of the distortion in the input audio signal, by applying the trained ML model to the image.
25. The method according to claim 24 , wherein the distortion comprises a Total Harmonic Distortion (THD).
26. The method according to claim 24 , wherein converting a given training audio signal into a given image comprises setting pixel values of the given image to represent an amplitude of the given training audio signal as a function of time.
27. The method according to claim 24 , wherein the respective images and the image are two-dimensional (2D).
28. The method according to claim 24 , wherein obtaining the training audio signals comprises (i) receiving initial audio signals having first durations, and (ii) slicing the initial audio signals into slices having second, shorter durations, so as to produce the training audio signals.
29. The method according to claim 24 , wherein the ML model comprises a convolutional neural network (CNN).
30. The method according to claim 24 , wherein the ML model comprises a generative adversary network (GAN).
31. The method according to claim 24 , wherein receiving the input audio signal comprises receiving the input audio signal from nonlinear audio processing circuitry.
32. The method according to claim 24 , wherein the ML model classifies the distortion according to the levels of distortion that label the training audio signal.
33. The method according to claim 24 , wherein the ML model estimates the level of distortion using regression.
34. The method according to claim 24 , and comprising controlling, using the estimated level of the distortion, an audio system that produces the input audio signal.
35. A method, comprising:
obtaining a plurality of initial audio signals, which have first durations in a first range of durations and which are labeled with respective levels of distortion;
slicing the initial audio signals into slices having second durations in a second range of durations, shorter than the first durations, so as to produce a set of training audio signals;
training a machine learning (ML) model to estimate the levels of the distortion based on the training audio signals, by (i) converting the training audio signals into respective images and (ii) training the ML model to estimate the levels of the distortion based on the images;
receiving an input audio signal having a duration in the second range of durations; and
estimating a level of the distortion in the input audio signal by applying the trained ML model to the input audio signal.
36. The method according to claim 35 , wherein the distortion comprises a Total Harmonic Distortion (THD).
37. The method according to claim 35 , wherein estimating the level of the distortion in the input audio signal comprises (i) converting the input audio signal into an image and (ii) applying the trained ML model to the image.
38. The method according to claim 35 , wherein the respective images are two-dimensional (2D) images.
39. The method according to claim 35 , wherein the ML model comprises a convolutional neural network (CNN).
40. The method according to claim 35 , wherein the ML model comprises a generative adversary network (GAN).
41. The method according to claim 35 , wherein the input audio signal is received from nonlinear audio processing circuitry.
42. The method according to claim 35 , wherein the ML model classifies the distortion according to the levels of distortion that label the training audio signal.
43. The method according to claim 35 , wherein the ML model estimates the level of distortion using regression.
44. The method according to claim 35 , and comprising controlling, using the estimated level of the distortion, an audio system that produces the input audio signal.