Method, apparatus and system for neural network hearing aid
The disclosure generally relates to a method, system and apparatus to improve a user's understanding of speech in real-time conversations by processing the audio through a neural network contained in a hearing device. The hearing device may be a headphone or hearing aid. In one embodiment, the disclosure relates to an apparatus to enhance incoming audio signal. The apparatus includes a controller to receive an incoming signal and provide a controller output signal; a neural network engine (NNE) circuitry in communication with the controller, the NNE circuitry activatable by the controller, the NNE circuitry configured to generate an NNE output signal from the controller output signal; and a digital signal processing (DSP) circuitry to receive one or more of controller output signal or the NNE circuitry output signal to thereby generate a processed signal; wherein the controller determines a processing path of the controller output signal through one of the DSP or the NNE circuitries as a function of one or more of predefined parameters, incoming signal characteristics and NNE circuitry feedback.
1. A hearing aid, comprising:
a microphone configured to receive a sound signal and output an electrical signal representing the sound signal;
a frontend receiver coupled to the microphone and configured to receive the electrical signal representing the sound signal, digitize the electrical signal, and output an audio signal comprising a digitized version of the sound signal;
an accelerometer configured to capture acceleration data for the hearing aid;
a voice activity detector configured to determine voice activity in the audio signal;
a sound level detector configured to determine whether a sound level of the audio signal exceeds a threshold sound level;
a first signal processing path and a second signal processing path, wherein the first signal processing path comprises neural network engine (NNE) circuitry configured to implement a neural network trained to perform background noise reduction, and the first signal processing path is configured to provide a higher target SNR enhancement than the second signal processing path; and
a controller configured to determine whether to transmit the audio signal to the first signal processing path or to the second signal processing path based on the acceleration data, the voice activity in the audio signal, and whether the sound level of the audio signal exceeds the threshold sound level.
2. The hearing aid of claim 1 , wherein the NNE circuitry is configured to perform at least 1 billion operations per second.
3. The hearing aid of claim 1 , wherein the NNE circuitry is configured to achieve at least 2 billion operations per milliwatt.
4. The hearing aid of claim 1 , wherein the NNE circuitry is configured to process the audio signal with an associated power consumption of about 2 milliwatts or less.
5. The hearing aid of claim 1 , wherein the NNE circuitry is configured to perform the background noise reduction based on a user selection of an operating mode through an application on a smartphone of the user.
6. The hearing aid of claim 1 , wherein the NNE circuitry is configured to perform the background noise reduction based on a user selection of an operating mode through an input on the hearing aid.
7. The hearing aid of claim 1 , wherein the NNE circuitry is configured to enhance sequentially-received signal samples of the audio signal and then output a processed signal as a continuous audio signal based on the enhanced sequentially-received signal samples in about 32 milliseconds or less of receipt of the audio signal.
8. The hearing aid of claim 1 , wherein the NNE circuitry is configured to obtain a speech component of the audio signal by:
estimating a complex ratio mask for the audio signal; or
subtracting a noise component of the audio signal from the audio signal.
9. The hearing aid of claim 1 , wherein the NNE circuitry is configured to obtain a noise component of the audio signal by:
estimating a complex ratio mask for the audio signal; or
subtracting a speech component of the audio signal from the audio signal.
10. The hearing aid of claim 1 , wherein the neural network comprises at least 1 million units.
11. A hearing aid, comprising:
a microphone configured to receive a sound signal and output an electrical signal representing the sound signal;
a frontend receiver coupled to the microphone and configured to receive the electrical signal representing the sound signal, digitize the electrical signal, and output an audio signal comprising a digitized version of the sound signal;
an integrated circuit (IC) comprising a first signal processing path, the first signal processing path comprising neural network engine (NNE) circuitry configured to implement a neural network trained to perform background noise reduction;
a second signal processing path not implemented on the IC, wherein:
the first signal processing path is configured to provide a higher target SNR enhancement than the second signal processing path; and
a controller configured to:
compute one or more metrics characterizing an acoustic environment; and
determine whether to transmit the audio signal to the first signal processing path or to the second signal processing path based on the computed one or more metrics.
12. The hearing aid of claim 11 , wherein the NNE circuitry is configured to perform at least 1 billion operations per second.
13. The hearing aid of claim 11 , wherein the NNE circuitry is configured to achieve at least 2 billion operations per milliwatt.
14. The hearing aid of claim 11 , wherein the NNE circuitry is configured to process the audio signal with an associated power consumption of about 2 milliwatts or less.
15. The hearing aid of claim 11 , wherein the NNE circuitry is configured to perform the background noise reduction based on a user selection of an operating mode through an application on a smartphone of the user.
16. The hearing aid of claim 11 , wherein the NNE circuitry is configured to perform the background noise reduction based on a user selection of an operating mode through an input on the hearing aid.
17. The hearing aid of claim 11 , wherein the NNE circuitry is configured to enhance sequentially-received signal samples of the audio signal and then output a processed signal as a continuous audio signal based on the enhanced sequentially-received signal samples in about 32 milliseconds or less of receipt of the audio signal.
18. The hearing aid of claim 11 , wherein the NNE circuitry is configured to obtain a speech component of the audio signal by:
estimating a complex ratio mask for the audio signal; or
subtracting a noise component of the audio signal from the audio signal.
19. The hearing aid of claim 11 , wherein the NNE circuitry is configured to obtain a noise component of the audio signal by:
estimating a complex ratio mask for the audio signal; or
subtracting a speech component of the audio signal from the audio signal.
20. The hearing aid of claim 11 , wherein the neural network comprises at least 1 million units.