Development and Analysis of Real-Time Acoustic Feedback Cancellation and a Clinical Testing Simulator for Hearing Studies
MetadataShow full item record
Acoustic feedback stems from acoustic coupling between the loudspeaker and microphone in hearing systems. Unpleasant outcomes of acoustic feedback, such as howling and whistling sounds, are major complaints of people with hearing impairment. As the size of hearing aid devices (HADs) have shrunk and the closed loop made by the feedback path has increased, acoustic feedback cancellation (AFC) methods have become more challenging for researchers. To overcome this problem, standard adaptive filters are commonly used to reduce the acoustic feedback. However, these methods are prone to biased estimations of the feedback path transfer function whenever the speech and feedback signals are correlated. The larger bias estimations worsen the performance of the system. The increased interest in the feedback cancellation problem has focused on the need for efficient AFC techniques. The noise injection method is one of many that is utilized to reduce the effect of acoustic feedback on HADs. This thesis proposes the development and analysis of real-time implementation of the noise injection method in changing room acoustics, a noise injection algorithm controlled by Spectral Flux feature-based voice activity detector (VAD). Objective evaluations were carried out in noisy conditions. Simulation results showed that the proposed methods are efficient in eliminating acoustic feedback. Additionally, real-time development and analysis for smartphone applications that can cancel acoustic feedback without distortion has been completed. In addition to feedback cancellation analysis, this work covers development of clinical testing software which is used in clinical testing stage to evaluate subjects’ hearing ability in noisy environments with various Signal-to-noise ratio (SNR) values.