Differences in fundamental frequency (F0) or pitch between competing voices facilitate our ability to segregate a target voice from interferers, thereby enhancing speech intelligibility. Although lower-numbered harmonics elicit a stronger and more accurate pitch sensation than higher-numbered harmonics, it is unclear whether the stronger pitch leads to an increased benefit of pitch differences when segregating competing talkers. To answer this question, sentence recognition was tested in young normal-hearing listeners in the presence of a single competing talker. The stimuli were presented in a broadband condition or were highpass or lowpass filtered to manipulate the pitch accuracy of the voicing, while maintaining roughly equal speech intelligibility in the highpass and lowpass regions. Performance was measured with average F0 differences (ΔF0) between the target and single-talker masker of 0, 2, and 4 semitones. Pitch discrimination abilities were also measured to confirm that the lowpass-filtered stimuli elicited greater pitch accuracy than the highpass-filtered stimuli. No interaction was found between filter type and ΔF0 in the sentence recognition task, suggesting little or no effect of harmonic rank or pitch accuracy on the ability to use F0 to segregate natural voices, even when the average ΔF0 is relatively small. The results suggest that listeners are able to obtain some benefit of pitch differences between competing voices, even when pitch salience and accuracy is low.
The accuracy with which we are able to discriminate the pitch of a harmonic complex tone depends on the F0 and the harmonic numbers present. For F0s in the average range of speech (100–200 Hz), pitch discrimination is best (implying accurate F0 coding) when harmonics below about the 10th are present [6–10]. When these lower-numbered harmonics are present, pitch discrimination is also independent of the phase relationships between the harmonics, suggesting that these harmonics are spectrally resolved to some extent. In contrast, when only harmonics above the 10th are present in this range of F0s, pitch discrimination is poorer and is affected by the phase relationships between harmonics, suggesting that interactions occur between these spectrally unresolved harmonics [6–10].
Psychoacoustic studies of sound segregation have often been carried out with interleaved sequences of tones. Some of these studies have investigated segregation based on differences in pitch accuracy and have varied the accuracy by systematically varying whether resolved or only unresolved harmonics are present. Previous studies have found that stream segregation can occur with alternating sequences of tones, even if the tones consist only of unresolved harmonics [11–14]. However, the question of whether streaming is greater with resolved than unresolved harmonics has received mixed answers. In cases where the listeners’ task was to segregate the streams, some studies have shown little difference in streaming between conditions containing resolved or only unresolved harmonics [11, 15], whereas another study using a similar approach found significantly greater stream segregation when resolved harmonics were present than when only unresolved harmonics were present [12]. However, in situations where the task was either neutral or encouraged listeners to integrate the sequences into a single stream, the results have been consistent across studies in showing greater segregation for complex tones containing resolved harmonics than for tones containing only unresolved harmonics [13, 14]. These findings support the idea that pitch accuracy can affect our ability to segregate sounds.
Less is known about the role of low-numbered harmonics in the context of segregating competing speech. Bird and Darwin [2] showed that lower harmonics dominate performance in a speech-segregation task based on F0 differences, but they did not test any conditions containing only high-numbered harmonics. Oxenham and Simonson [16] explored the effect of harmonic rank on speech intelligibility by comparing conditions where the target and single-talker masker had been lowpass (LP) or highpass (HP) filtered to either retain (LP-filtered) or remove (HP-filtered) the spectrally resolved components from the target and masker [16]. The LP and HP cutoff frequencies were selected to produce roughly equal performance in noise for both conditions. Surprisingly, performance in the LP and HP conditions improved by similar amounts when the noise masker was replaced by a single-talker masker with a different average F0, suggesting no clear benefit of having resolved harmonic components in the speech. However, that study only used relatively large values of average ΔF0 that according to recent F0 estimates were approximately 4 and 8 semitones (ST). Moreover, this study did not parametrically vary the ΔF0 between the target and masker. It may be that pitch accuracy is only relevant for more challenging conditions, i.e. for conditions with smaller average values of ΔF0. Thus, it remains unclear whether the effect of ΔF0 on performance is affected by the presence or absence of low-numbered, spectrally resolved harmonics. The aim of the present study was to determine whether there is an effect of spectral region, and hence pitch coding accuracy, on the ability of listeners to use average F0 differences between a target and an interfering talker to understand natural speech.
Pitch discrimination is better for synthetic timbre than natural musical instrument timbres, despite familiarity
