Method, apparatus and system for neural network hearing aid

An ear-worn device for audio enhancement, the ear-worn device comprising:

• neural network engine (NNE) circuitry configured to receive an audio signal and generate a plurality of intermediate signals, the plurality of intermediate signals comprising a speech signal and a noise signal;
• performance measurement circuitry configured to determine a performance metric based on an output of the NNE circuitry;
• and
• recombiner circuitry configured to receive the plurality of intermediate signals and combine the speech signal and the noise signal into an output signal, wherein respective gains applied to the speech signal and the noise signal are determined based on the performance metric.

An ear-worn device for audio enhancement, the ear-worn device comprising:

• neural network engine (NNE) circuitry configured to generate an output signal from an incoming signal using neural network processing;
• performance measurement circuitry configured to determine a performance metric based on an output of the NNE circuitry, wherein the performance metric predicts performance of the NNE circuitry;
• and
• decision logic circuitry configured to determine whether to turn the neural network processing off based on the performance metric.