Deep neural network based environment sound classification and its implementation on hearing aid app.

• Using 5 kinds of environment audio data collected by mobile phone. • The sound environment classification system stacking 7 audio frames into a block. • The sound environment classification system integrating 30 rows into an audio unit. • Transplantation of sound environment classification system on mobile phone. In general, a hearing aid app is very useful for the persons having either partial or complete inability to hear. At present, there is no special provision available in the hearing aid app for the classification of different environmental sounds. This paper proposes an algorithm for environmental sound classification based on Superimposed Audio Blocks using Deep Neural Networks (SAB - DNN) and also to implement it on the hearing aid app. The system can recognize five kinds of different sound fields automatically: bus, subway, street, indoor, car. In this system, 512 sampling points are taken as an audio frame and several audio frames are stacked up into an Audio Block (AB). when 7 audio frames are stacked up into an Audio Block (AB), the accuracy rate of sound environment classification using (AB - DNN) tends to be the best (96.18%). Based on this, the experiment integrates multiple Audio Block (AB) into an audio unit called Superimposed Audio Blocks(SAB) and classify it using DNN. Optimally, 30 sound blocks are integrated into a SAB which results in the classification accuracy up to 98.8%. As far as we know, it is the first time on the hearing aid app to implement an improved Deep Neural Network (DNN) based classification system and superposition of multi-audio frames and blocks. [ABSTRACT FROM AUTHOR]