Adapting Automatic Speech Recognition to In-Ear Microphone Speech

Traditionally, speech is captured from in front of the mouth. Existing an Automatic Speech Recognition (ASR) are all made to work with this type of speech. However, noise and other factors can be detrimental to the performance of ASR systems. When using an in-ear device that blocks the earcanal, a microphone can be placed inside the ear to capture speech that is relatively not affected by noise. Speech captured from inside the ear sounds different from speech captured in front of the mouth. It has a lot more low frequency content and not enough high frequency content. The aim of this project is to develop an ASR that is adapted to in-ear microphone speech. For this work, it will be implemented on an advanced earpiece developed by EERS to work with an e-bike developed by Nerra.

Faculty Supervisor:

Rachel Bouserhal

Student:

Partner:

EERS Global Technologies Inc

Discipline:

Engineering

Sector:

Manufacturing; Professional, scientific and technical services

University:

École de technologie supérieure

Program: