A new feature selection method for volume control in direct-learning hearing aid systems

Jin Zhou, Hisham Othman, Hilmi Dajani, Tyseer Aboulnasr

Abstract


A study was conducted to demonstrate an artificial neural network (ANN) as a new feature selection method for volume control in direct-learning hearing aid systems. The goal of the study was to derive suitable features that the ANN used to set the volume such that it optimized speech intelligibility. New features were proposed that were based on measures of speech intelligibility, such as the Speech Intelligibility Index (SII) and the Coherence SII. The performance of these features was investigated using a simulator of a hearing aid user. These features were derived from the calculation of the CII in the third-octave bands and they reflected the SNR and energy of the speech signals in the bands weighted by the psychoacoustic characteristics of the listener.

Keywords


Audition; Backpropagation; Gain control; Hearing aids; Neural networks; Artificial Neural Network; Feature selection methods; Hearing aid; Octave bands; Speech intelligibility index; Speech signals; Volume controls

Full Text:

PDF

Refbacks

  • There are currently no refbacks.