Speech intelligibility and talker gender classification with noise-vocoded and tone-vocoded speech.

Sarah Villard,Gerald Kidd

doi:10.1121/10.0006285

Abstract

Vocoded speech provides less spectral information than natural, unprocessed speech, negatively affecting listener performance on speech intelligibility and talker gender classification tasks. In this study, young normal-hearing participants listened to noise-vocoded and tone-vocoded (i.e., sinewave-vocoded) sentences containing 1, 2, 4, 8, 16, or 32 channels, as well as non-vocoded sentences, and reported the words heard as well as the gender of the talker. Overall, performance was significantly better with tone-vocoded than noise-vocoded speech for both tasks. Within the talker gender classification task, biases in performance were observed for lower numbers of channels, especially when using the noise carrier.

Highlights

Since the development of the vocoder in the first half of the 20th century (Dudley, 1939), many scientific pursuits and clinical applications have arisen that were based on the underlying technique; that is, imposing the amplitude envelopes derived on a band-by-band basis from a broadband signal on a specified carrier
There is a wide range of applications of vocoding in modern communications and music synthesis technology, the most common topic areas in the speech perception and psychological acoustics fields involve using vocoding to control various aspects of a signal in the study of normal auditory processes and as a useful analog to the information conveyed by cochlear implants (CIs), especially with regard to speech recognition (Shannon et al, 1995)
Recent work has established that differences in talker gender can be a powerful cue to overcoming masking in multiple-talker listening situations (e.g., Kidd et al, 2016) and that the control of the stimulus parameters that is possible via vocoding can be useful in determining the different factors that contribute to speech-on-speech masking (e.g., Arbogast et al, 2002; Swaminathan et al, 2016)

Summary

Introduction

Since the development of the vocoder in the first half of the 20th century (Dudley, 1939), many scientific pursuits and clinical applications have arisen that were based on the underlying technique; that is, imposing the amplitude envelopes derived on a band-by-band basis from a broadband signal (e.g., speech or music) on a specified carrier (e.g., continuous noise, tones, or pulse trains). There is a wide range of applications of vocoding in modern communications and music synthesis technology, the most common topic areas in the speech perception and psychological acoustics fields involve using vocoding to control various aspects of a signal in the study of normal auditory processes (e.g., pitch, binaural hearing, masking) and as a useful ( limited) analog to the information conveyed by cochlear implants (CIs), especially with regard to speech recognition (Shannon et al, 1995). Recent work has established that differences in talker gender can be a powerful cue to overcoming masking in multiple-talker listening situations (e.g., Kidd et al, 2016) and that the control of the stimulus parameters that is possible via vocoding can be useful in determining the different factors that contribute to speech-on-speech masking (e.g., Arbogast et al, 2002; Swaminathan et al, 2016). It is not our intent to draw conclusions about the perceptual abilities of CI listeners by using vocoder simulations in normal-hearing (NH) listeners, we do hope to establish a baseline against which future studies with CI listeners may be compared, in keeping with many past studies that have contributed to this body of knowledge (e.g., Fu et al, 2004; Whitmal et al, 2007; Yun et al, 2021)

Methods

Results

Conclusion