Real-time Signal Processing Algorithms

Cochlear implant (CI) patients have difficultyunderstanding tonal languages which use pitchvariations to convey meaning because of poor pitchperception. The ability of CI patients to communicatein noisy environments is severely degraded.Therefore pitch-detection and (spatial) noisesuppression is of importance to current cochlearimplant devices. This thesis provides the details ofreal-time implementations that address theseconcerns. A pitch-detector was implemented toestimate the pitch and a beamformer was implementedto suppress spatial noise. The pitch-detection andthe beamforming algorithms were implemented on an ARMbased processor of a Personal Digital Assistant(PDA). The pitch-detection algorithm is based on theautocorrelation function. Its real-time performancewas measured and its efficiency in the presence ofspeech-shaped noise and babble noise was evaluated.The beamforming algorithm implemented was theGriffith's beamformer. Objective and subjective testsfor different signal-to-noise ratio levels ofspeech-shaped noise were conducted to evaluate itsperformance. This thesis includes the results ofthese evaluations.