Multiband equalization tuning and control based on artificial intelligence
One embodiment provides a computer-implemented method that includes accessing an artificial intelligence model trained for a filterbank based on a control gain of the filterbank and a resulting frequency response gain. Based on a target frequency response gain inputted into the trained artificial intelligence model, a control gain is applicable to a filter in the filterbank is outputted. The target frequency response gain is obtained at a center frequency of the filter in the filterbank.
1 . A computer-implemented method comprising:
accessing an artificial intelligence model trained, using training data comprising pairs of control gain vectors and measured frequency response gains produced by a filterbank, to provide a mapping from a target frequency response gain vector to a corresponding vector of control gains for the filterbank;
inputting a target frequency response gain vector into the trained artificial intelligence model;
outputting, from a single inference of the trained artificial intelligence model and without iterative tuning, a vector of control gains respectively applicable to filters of the filterbank; and
applying the vector of control gains to the filterbank to obtain frequency response gains at center frequencies of the filters that correspond to the target frequency response gain.
2 . The computer-implemented method of claim 1 , wherein the trained artificial intelligence model develops a learned relationship between the control gain vector and the resulting frequency response gain vector for the filterbank, and the filterbank is part of a multiband graphic equalizer.
3 . The computer-implemented method of claim 1 , wherein applying the vector of control gains produces output frequency response gains that match the target frequency response gain vector within an allowable deviation.
4 . The computer-implemented method of claim 1 , wherein the artificial intelligence model comprises a neural network.
5 . The computer-implemented method of claim 4 , wherein the training data comprises random control gain vectors generated using a uniform distribution and their corresponding measured frequency response gains obtained from the filterbank.
6 . The computer-implemented method of claim 4 , wherein the neural network is trained to adjust control gains of N filters to control target frequency response gains at M points, N and M are integers, and M is greater than N.
7 . The computer-implemented method of claim 4 , wherein the neural network is further configured to adjust coefficients of a set of biquad filters associated with the filterbank to obtain the target frequency response gain vector, and the adjustments of the coefficients are determined without iterative tuning to achieve the target frequency response gain across varying audio signal conditions.
8 . A non-transitory processor-readable medium that includes a program that when executed by a processor performs applying a vector of control gains to a filterbank using a trained artificial intelligence model, comprising:
accessing, by the processor, an artificial intelligence model trained, using training data comprising pairs of control gain vectors and measured frequency response gains produced by a filterbank, to provide a mapping from a target frequency response gain vector to a corresponding vector of control gains for the filterbank;
inputting, by the processor, a target frequency response gain vector into the trained artificial intelligence model;
outputting, by the processor, from a single inference of the trained artificial intelligence model and without iterative tuning, a vector of control gains respectively applicable to filters of the filterbank; and
applying, by the processor, the vector of control gains to the filterbank to obtain frequency response gains at center frequencies of the filters that correspond the target frequency response gain.
9 . The non-transitory processor-readable medium of claim 8 , wherein the trained artificial intelligence model develops a learned relationship between the control gain vector and the resulting frequency response gain vector for the filterbank, and the filterbank is part of a multiband graphic equalizer.
10 . The non-transitory processor-readable medium of claim 8 , wherein applying the vector of control gains produces output frequency response gains that match the target frequency response gain vector within an allowable deviation.
11 . The non-transitory processor-readable medium of claim 8 , wherein the artificial intelligence model comprises a neural network.
12 . The non-transitory processor-readable medium of claim 11 , wherein the training data comprises random control gain vectors generated using a uniform distribution and their corresponding measured frequency-response gains obtained from the filterbank.
13 . The non-transitory processor-readable medium of claim 11 , wherein the neural network is trained to adjust control gains of N filters to control target frequency response gains at M points, N and M are integers, and M is greater than N.
14 . The non-transitory processor-readable medium of claim 11 , wherein the neural network is further configured to adjust coefficients of a set of biquad filters associated with the filterbank to obtain the target frequency response gain vector, and the adjustments of coefficients are determined without iterative tuning to achieve the target frequency response gain across varying audio signal conditions.
15 . An apparatus comprising:
a memory storing instructions; and
at least one processor executes the instructions including a process configured to:
access an artificial intelligence model trained, using training data comprising pairs of control gain vectors and measured frequency response gains produced by a filterbank, to provide a mapping from a target frequency response gain vector to a corresponding vector of control gains for the filterbank;
input a target frequency response gain vector into the trained artificial intelligence model;
output, from a single inference of the trained artificial intelligence model and without iterative tuning, a vector of control gains respectively applicable to filters of the filterbank; and
apply the vector control gains to the filterbank to obtain frequency response gains at center frequencies of the filters that correspond to the target frequency response gain.
16 . The apparatus of claim 15 , wherein the trained artificial intelligence model develops a learned relationship between the control gain vector and the resulting frequency response gain vector for the filterbank, and the filterbank is part of a multiband graphic equalizer.
17 . The apparatus of claim 15 , wherein applying the vector of control gains produces output frequency response gains that match the target frequency response gain vector within an allowable deviation.
18 . The apparatus of claim 15 , wherein the artificial intelligence model comprises a neural network, and the training data comprises random control gain vectors generated using a uniform distribution and their corresponding measured frequency response gains obtained from the filterbank.
19 . The apparatus of claim 18 , wherein the neural network is trained to adjust control gains of N filters to control target frequency response gains at M points, N and M are integers, and M is greater than N.
20 . The apparatus of claim 18 , wherein the neural network is further configured to adjust coefficients of a set of biquad filters associated with the filterbank to obtain the target frequency response gain vector, and the adjustments of the coefficients are determined without iterative tuning to achieve the target frequency response gain across varying audio signal conditions.