Effects of electrode separation between speech and noise signals on consonant identification in cochlear implants

Abstract

The aim of the present study was to examine cochlear implant ͑CI͒ users' perceptual segregation of speech from background noise with differing degrees of electrode separation between speech and noise. Eleven users of the nucleus CI system were tested on consonant identification using an experimental processing scheme called "multi-stream processing" in which speech and noise stimuli were processed separately and interleaved. Speech was presented to either ten ͑every other electrode͒ or six electrodes ͑every fourth electrode͒. Noise was routed to either the same ͑the "overlapped" condition͒ or a different set of electrodes ͑the "interlaced" condition͒, where speech and noise electrodes were separated by one-and two-electrode spacings for ten-and six-electrode presentations, respectively. Results indicated a small but significant improvement in consonant recognition ͑5%-10%͒ in the interlaced condition with a two-electrode spacing ͑approximately 1.1 mm͒ in two subjects. It appears that the results were influenced by peripheral channel interactions, partially accounting for individual variability. Although the overall effect was small and observed from a small number of subjects, the present study demonstrated that CI users' performance on segregating the target from the background might be improved if these sounds were presented with sufficient peripheral separation.

Identification and Multiplicity of Double Vowels in Cochlear Implant Users

“…As discussed earlier, the method of interleaving the two streams in the present study was also different: Although intimate understanding of the technical difference is not critical to understanding the rationale of the present study, interested readers are encouraged to compare Figure 1 in the present article to Figure 1 of the previous article (Kwon, 2009). Table 4 displays percent scores of vowel identification in the Single session for each subject.…”

“…Therefore, stimulus pulses for MSP were constructed with an experimental map of 250 pulses per second (pps) per channel, which was obtained separately prior to data collection. In MSP, pulse streams from the two vowels were interleaved in blocks of eight pulses, rather than interleaving pulse by pulse as done in the previous study (Kwon, 2009), in order to reduce possible channel crosstalk between the two pulse streams (cf. Figure 1).…”

“…For MSP, unlike in the previous study (Kwon, 2009), the number of maxima was eight, and all electrode channels were used in ACE™ processing (i.e., no preselected channels such as odd-or even-numbered electrodes). As discussed earlier, the method of interleaving the two streams in the present study was also different: Although intimate understanding of the technical difference is not critical to understanding the rationale of the present study, interested readers are encouraged to compare Figure 1 in the present article to Figure 1 of the previous article (Kwon, 2009).…”

“…Stimulation pulse preparation for the MSP condition, signal processing parameters, and the method of pulse stream arrangement were chosen to maximize temporal separation of the two pulse streams based on observations in the previous study (Kwon, 2009). Therefore, stimulus pulses for MSP were constructed with an experimental map of 250 pulses per second (pps) per channel, which was obtained separately prior to data collection.…”

“…As a result, the pattern of electrical stimulation for the combined signal can differ, sometimes substantially, from the electrical stimulation that would be produced from either of the two constituent sounds alone. Kwon (2009) argued that, although studies based on commonplace CI strategies might appropriately demonstrate the user's realistic performance, such studies likely overlook fundamental insight into auditory grouping (or the lack thereof ) in CIs. As an alternative, Kwon proposed that the acoustic-to-electric conversion be done individually for each sound and the pulse sequences from the two sounds be interleaved in time for the delivery of the stimulus.…”