Real-Time Hardware Implementation of Telephone Speech Enhancement Algorithm

Abstract

Hearing impairment detrimentally affects communication over the telephone. Since phone lines reduce bandwidth and dynamic range, the poor quality speech signal can cause hard of hearing (HoH) listeners to experience extreme frustration and inefficient communication. One possible solution has been developed at the Ohio State University to help combat this problem. The Telephone Speech Enhancement Algorithm (TSEA) has been created to improve telephone signals so that speech is more intelligible for HoH listeners. Tests for TSEA have been run on human subjects and proven the algorithm effective. However, a hardware implementation of TSEA has yet to be designed. In this thesis, the BeagleBoard-xM development board is used to run TSEA. The software for TSEA is modified so that it can be implemented on the BeagleBoard-xM and tested in a real-time environment. This hardware model runs TSEA but introduces noise into the system due to its analog nature. The model accepts analog audio signals, processes them using TSEA, and outputs the processed signal for transmission. A device such as this has the potential to improve communication in scenarios such as telemedicine clinics where a failure to communicate properly with their HoH customers could have potentially devastating consequences. Ideally if a commercial model was developed, TSEA could be implemented everywhere to help improve communications for the HoH community. This project is the next step in making it a reality.