Page 29 - Spring 2015
P. 29
high levels of sound too often, or simply ages. Hair cells in mammals do not regrow when they die, meaning that if we want to pass acoustic information to the brain in the absence of usable hair cells, we need a way to bypass the hair cell transduction process, such as with a cochlear implant (CI).
The modern-day CI is a biomedical device that includes a microphone on a speech processor placed behind the ear and an array of electrodes that are placed in the cochlea (Figure 1). Frequencies most important for speech under- standing (typically about 200-8,000 Hz) are passed to the auditory system through the electrodes by their direct ex- citation of the spiral ganglia attached to the auditory nerve fibers, thereby bypassing the dead hair cells. Looked at an- other way, stimulation by the CI starts the electrochemical transduction process in the nerve with the electricity rather than the normal chemical transfer from the hair cells. Each electrode contact is assigned a frequency range, and the fre- quency-to-electrode allocation is such that the organization of frequencies follow that of the functional cochlea, high frequencies near the bottom and low frequencies near the top. Therefore, the CI mimics the frequency organization of the cochlea to maintain its ability to operate as a frequency analyzer.
Today’s CIs are marvels of modern biomedical engineering and the most successful of all sensory prostheses. In the best cases, people with CIs achieve near-perfect speech under- standing in quiet conditions and can speak on the phone, thereby understanding degraded speech signals without the help of visual cues. The pioneers of the modern-day mul- tielectrode CI (Graeme M. Clark, Emeritus, University of Melbourne; Ingeborg Hochmair, MED-EL, Innsbruck; and Blake S. Wilson, Duke University) were recently recognized with the Lasker Award “for the development of the modern CI — a device that bestows hearing to individuals with pro- found deafness” (Pierce, 2013).
The National Institute on Deafness and Other Communi- cation Disorders (NIDCD) of the National Institutes of Health (NIH) approximates that there are over 324,000 CI users worldwide (NIDCD, 2012). These users range over the full life span, with children being implanted at younger and younger ages (sometimes at six or nine months of age, which works because the cochlea is nearly full size when born and grows little as one ages) and adults at older and older ages (sometimes at 90+ years).
Why undergo the risk of major surgery for people with relatively fragile constitutions such as infants and seniors? People receive CIs when quality of life is an issue. Humans are social beings and without the ability to communicate with each other, many become reclusive and depressed. A majority of the world relies on hearing to communicate and socialize. In addition, many have established social circles, friends, and family for which communication stops with the loss of hearing, even in the presence of communication al- ternatives.
CI speech-Processing basics
Arguably, the last major advance in CI signal processing was over 20 years ago (Wilson et al., 1991) with the introduc- tion of multichannel high carrier rate strategies (meaning that the stimulation rate on an individual electrode is closer to 1,000 pulses/s rather than the 100 pulses/s in the earliest CIs). Today’s modern implants have between 12 and 22 in- tracochlear electrodes that serve as information channels to convey temporal envelope information (the slowly varying temporal information in a signal rather than the fast oscil- lations of the carrier or fine structure; see Figure 2). After all the filtering from microphone responses, preprocessing, and compression, all CIs operate using “vocoder-centric” signal processing (Loizou, 2006), which is based on a meth- od of efficient information encoding originally designed by Dudley (1959). This partitions the spectrum into informa- tion channels, providing worse resolution than those with typical acoustic hearing (see Figure 2, Analysis Stage) but which, as discussed below, is sufficient for speech under- standing. People with CIs receive on the order of 10 discrete spectral channels, whereas the typical auditory system can be best viewed as a continuum of auditory filters with orders of magnitude finer spectral resolution. Comparing typical hearing with CIs is like comparing modern-day graphing calculators with the old mechanical calculators. If all you knew were mechanical calculators, the modern-day graph- ing calculators that could compute your logarithms for you would seem grand and wonderful. However, no one who grew up on modern-day graphing calculators would tolerate a mechanical calculator and compute their logarithms with the help of a table unless they had to.
The next stage of CI processing is the extraction of the tem- poral envelope from each channel. The exact extraction method varies in practice, but the idea is the same in all cases. What matters is the slowly varying amplitude modu-
| 27