Real-time implementation of signal processing algorithms for cochlear implant applications

Cochlear implant (CI) patients have difficulty understanding tonal languages which make use of pitch variations to convey meaning. The ability of CI patients to communicate in noisy environments is severely degraded. Therefore, pitch-detection and (spatial) noise suppression are of importance to current cochlear implant devices. This thesis provides the details of real-time implementations of algorithms that address these concerns. A pitch-detector was implemented to estimate the pitch and a beamformer was implemented to suppress noise in situations where noise is spatially separated from the target speaker. The pitch-detection and the beamforming algorithms were implemented on an ARM based processor of a Personal Digital Assistant (PDA). The pitch-detection algorithm is based on the autocorrelation function. Its real-time performance was measured and its efficiency in the presence of speech-shaped noise and babble noise was evaluated. The Griffiths-Jim's beamforming algorithm was implemented. Objective and subjective listening tests with normal-hearing listeners were conducted to evaluate its performance. This thesis presents the results of these evaluations.