Low-frequency speech cues and simulated electric-acoustic hearing

Abstract

The addition of low-frequency acoustic information to real or simulated electric stimulation (so-called electric-acoustic stimulation or EAS) often results in large improvements in intelligibility, particularly in competing backgrounds. This may reflect the availability of fundamental frequency (F0) information in the acoustic region. The contributions of F0 and the amplitude envelope (as well as voicing) of speech to simulated EAS was examined by replacing the low-frequency speech with a tone that was modulated in frequency to track the F0 of the speech, in amplitude with the envelope of the low-frequency speech, or both. A four-channel vocoder simulated electric hearing. Significant benefit over vocoder alone was observed with the addition of a tone carrying F0 or envelope cues, and both cues combined typically provided significantly more benefit than either alone. The intelligibility improvement over vocoder was between 24 and 57 percentage points, and was unaffected by the presence of a tone carrying these cues from a background talker. These results confirm the importance of the F0 of target speech for EAS (in simulation). They indicate that significant benefit can be provided by a tone carrying F0 and amplitude envelope cues. The results support a glimpsing account of EAS and argue against segregation.

Figure 1

Mean percent correct scores, ±1 standard error. Each curve represents a different background, which was present in the vocoder region only. Processing conditions are along the x axis. The output of the vocoder was presented alone (V) or with different low-frequency cues, which included 500 Hz low-pass speech (V∕500), a tone with only voicing applied (V∕T), a tone modulated in frequency by the fundamental of the target talker (V∕T_F0), a tone modulated in amplitude by the envelope of the low-pass speech (V∕T_env), and a tone modulated in both frequency and amplitude (V∕T_F0-env). Mean intelligibility provided by the low-frequency cues themselves is depicted by diamonds. Values to the right of the V∕T_F0-env data points are percentage points of improvement over vocoder alone. Although the processing variable, which is depicted along the x axis, is not a continuous variable, the different levels of processing within each background are connected with lines for clarity.

Figure 2

Mean percent correct scores, ±1 standard error. Squares and circles indicate performance when the background was a male and a female, respectively. Unfilled and filled symbols indicate performance when the low-frequency cue was speech and a modulated tone, respectively. Different low-frequency stimulus conditions are represented along the x axis, and are no low-frequency stimulus (none), target only (target), background only (background), or target and background (target+background). Although the low-frequency stimulus variable, which is depicted along the x axis, is not a continuous variable, the different levels of this variable within each background gender∕low-frequency cue combination are connected with lines for clarity.