Quantitative Analysis of Neural Systems: An Analysis of Parameter Variations in Cochlear Implant Performance Jorge Roldan Dept. of Biomedical Engineering Louisiana Tech University November 14, 2002 Cochlear Implants • Medical device designed to electrically stimulate the auditory nerve • Main features – Speech Processors • Speech detected by microphones and sent to processors • Signals are filtered, analyzed, and digitized – Internal components • Signals transmitted via FM signals to implant • Corresponding current delivered through electrode – Speech Coding Strategies • Controlled the way signal is processed • Emphasize different qualities such as pitch and loudness • SPEAK – Spectral Peak Coding Strategy • CIS – Continuous Interleaved Sampling Physiology • Auditory Nerve System – Comprised of central and peripheral systems • Peripheral – converts sound into neural codes – External, middle and inner ear • Central – interprets signals as speech patterns, music, noise etc. – Inner Ear and cochlea • Transduces mechanical forces (fluid waves) to electrical signals via cilia – Frequency Spectrum • Fundamental Frequency (vowels) • Harmonics vary across frequency indicating resonances (formats) Implant Parameters I • Number of implanted electrodes – Can vary from 1 to 22 – 7 electrodes sufficient for speech perception – Greater number allow for different configurations, frequency allocations – flexibility to eliminate faulty electrodes or those that produce unwanted sensations • Electrode configuration – Refers to location, orientation, and separation • Best performance at apical location • 3.75 mm distance more closely approximates normal frequency delivery • Minimizes electrode interaction and interference • Monopolar vs. Bipolar – Bipolar should be more specific – Monopolar offer broad stimulation Implant Parameters II • Frequency bandwidth and allocation – Signal bandwidth can be no higher than half the sampling frequency – Normal human tonal frequencies • 20 Hz – 20 kHz • Only 3kHz needed to detect first 3 formats • f >6.7 kHz produces no improved effect on speech recognition • Rate and duration of stimulus pulses – Pulse rate determined by pulse width (vice versa) – Faster rates (>800 pps) offer greatest consonant recognition – Long pulse widths offer greater dynamic range when using slow rates Implant Parameters III • Amplitude Mapping Functions – Transformation of acoustic amplitude to electrical amplitude – Normal speech sounds range from 40 – 60 dB – Implants listeners have dynamic range from 6-20 dB – Power-Law or Logarithmic compression • Varying Exponent Conclusions • Electrodes sufficient for perception can be low but greater numbers offer flexibility • High pulse rates with small width provide best performance • Bandwidth above 6.7 kHz show no significant effect • Frequency allocation should be optimized for given electrode configuration • Power-law transformations have no effect on speech perception. Logarithmic transformations work well BEAGLE MEAT It’s what’s for DINNER !!!
Pages to are hidden for
"Quantitative Analysis of Neural Systems An Analysis of Parameter"Please download to view full document