On the Training of Multilayer Perceptrons for Speech/Non-Speech Classification in Hearing Aids
This paper explores the application of multilayer perceptrons (MLP) to the problem of speech/non-speech classification in digital hearing aids. When properly designed and trained, MLPs are able to generate an arbitrary classification frontier with a relatively low computational complexity. The paper will focus on studying the key influence of the training process on the performance of the system. An appropriate election of the training algorithm will help to provide better classification with a lower number of neurons in the network, which leads to a lower computational complexity. The results obtained will be compared with those obtained from two reference algorithms (the Fisher linear discriminant and the k-Nearest Neighbour), along with some comments regarding the computational complexity.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is temporarily free for AES members.