EVALUATION OF A NEW SPECTRAL PEAK CODING STRATEGY FOR THE NUCLEUS 22 CHANNEL COCHLEAR IMPLANT SYSTEM

Sixty-three postlinguistically deaf adults from four English-speaking countries participated in a 17-week field study of performance with a new speech coding strategy, Spectral Peak (SPEAK), and the most widely used strategy, Multipeak (MPEAK), both of which are implemented on we arable speech processors of the Nucleus 22 Channel Cochlear Implant Sy stem; MPEAK is a feature-extraction strategy, whereas SPEAK is a filte rbank strategy. Subjects' performance was evaluated with an experiment al design in which use of each strategy was reversed and replicated (A BAB). Average scores for speech tests presented sound-only at 70 dB SP L were higher with the SPEAK strategy than with the MPEAK strategy. Fo r tests in quiet, mean scores for medial vowels were 74.8 percent vers us 70.1 percent; for medial consonants, 68.6 percent versus 56.6 perce nt; for monosyllabic words, 33.8 percent versus 24.6 percent; and for sentences, 77.5 percent versus 67.4 percent. For tests in noise, mean scores for Four-Choice Spondees at +10 and +5 dB signal-to-noise ratio (S/N) were 88.5 percent versus 73.6 percent and 80.1 percent versus 6 2.3 percent, respectively; and for sentences at +15 dB, +10, and +5 dB S/N, 66.5 percent versus 43.4 percent, 61.5 percent versus 37.1 perce nt, and 60.4 percent versus 31.7 percent, respectively. Subjects showe d marked improvement in recognition of sentences in noise with the new SPEAK filterbank strategy. These results agree closely with subjects' responses to a questionnaire on which approximately 80 percent report ed they heard best with the SPEAK strategy for everyday listening situ ations.