Implementation and Tuning of Multi-rate Speech Processors for Cochlear Implants

点击翻译按钮获取中文摘要

A variety of approaches have been considered to improve speech recognition by cochlear implant subjects over the last 20 years. One approach involves using experimental data that probes the status of a particular subjects' electrically stimulated auditory system to tune conventional speech processing systems. Another approach involves investigating new speech processing approaches and algorithms. The goal of the proposed research is to link these two successful research methodologies by investigating the feasibility and optimization of novel speech processing techniques by specifically tuning the algorithm to the measured psychophysical parameters associated with a particular implant subject. The hypothesis that will be investigated is whether encoding dynamic changes in the frequency information present in speech by specifically encoding this information in a variable electrical pulse rate can improve speech understanding by cochlear implant subjects. In the first Specific Aim, a finite set of pulse rates, each based on psychophysical data collected from the subject being tested, will be used to code speech information in each electrode channel. This differs from traditional speech processing techniques which use a non-tuned fixed pulse rate in each channel, and differs from recently presented speech processing strategies in that a discrete number of rates, as opposed to a continuum, are utilized. The second Specific Aim will consider the utility of coding speech in virtual channels, where each virtual channel is defined by an electrode and a stimulation rate. Both Aims will be addressed by testing the proposed speech processing algorithms after adjusting the parameters of the algorithms based on measured psychophysical data. In the United States alone, there are 28 million individuals with hearing loss, and 1-2 million completely deaf individuals. As a result of increasing levels of noise in the modern environment, hearing loss ocurs earlier in life and progresses more rapidly, suggesting that these numbers will continue to grow. The goal of this research is to provide improved speech recognition performance for such individuals by optimizing novel techniques for processing and coding speech information.