Harisadhan Patra

The role of suppression in psychophysical tone-on-tone masking

This study tested the hypothesis that suppression contributes to the difference between simultaneous masking (SM) and forward masking (FM). To obtain an alternative estimate of suppression, distortion-product otoacoustic emissions (DPOAEs) were measured in the presence of a suppressor tone. Psychophysical-masking and DPOAE-suppression measurements were made in 22 normal-hearing subjects for a 4000-Hz signal/f(2) and two masker/suppressor frequencies: 2141 and 4281 Hz. Differences between SM and FM at the same masker level were used to provide a psychophysical estimate of suppression. The increase in L(2) to maintain a constant output (L(d)) provided a DPOAE estimate of suppression for a range of suppressor levels. The similarity of the psychophysical and DPOAE estimates for the two masker/suppressor frequencies suggests that the difference in amount of masking between SM and FM is at least partially due to suppression.

Relative effects of increment and pedestal duration on the detection of intensity increments.

The detection of a brief increment in the intensity of a longer duration pedestal is commonly used as a measure of intensity-resolution. Increment detection is known to improve with increasing duration of the increment and also with increasing duration of the pedestal, but the relative effects of these two parameters have not been explored in the same study. In several past studies of the effects of increment duration, pedestal duration was increased as increment duration increased. In the present study, increment and pedestal duration were independently manipulated. Increment-detection thresholds were determined for four subjects with normal-hearing using a 500- or 4000-Hz pedestal presented at 60 dB sound pressure level (SPL). Increment durations were 10, 20, 40, 80, 160, and 320 ms. Pedestal durations were 20, 40, 80, 160, and 320 ms. Each increment duration was combined with all pedestals of equal or greater duration. Multiple-regression analyses indicate that increment detection under these conditions is determined primarily by pedestal duration. Follow-up experiments ruled out effects of off-frequency listening or overshoot. The results suggest that effects of increment duration have been confounded by effects of pedestal duration in studies that co-varied increment and pedestal duration. Implications for models of temporal integration are discussed.

Effect of narrowband noise maskers on increment detection.

The ability to detect a brief increment in intensity in a longer duration stimulus is often used as a measure of the intensity resolution of the auditory system, but it is unclear how listeners perform the task. Early investigators suggested that detection of an increment in intensity was based on the overall energy within the critical band centered on the signal frequency. The use of signal energy falling outside of the critical band is often limited by introducing a noise masker. Recent findings indicate that envelope fluctuations in a masker can increase thresholds in increment detection, suggesting a decision process based on envelope modulation. A modulation‐based decision device was evaluated by adding irregularity to the envelope of an increment detection task via noise maskers of varying bandwidth. Results indicate that thresholds for detection of an increment in the intensity of a pure tone increase with increasing bandwidth for an on‐frequency noise masker, but are unchanged by an off‐frequency ...

The relative weighting of energy and envelope‐modulation cues in increment detection.

Recent findings suggest that listeners primarily rely on envelope‐modulation cues rather than energy cues in an increment detection task, where one of the two sounds contains a brief increment in intensity in a longer duration tone. The relative contributions of energy and envelope‐modulation cues in increment detection were evaluated in six normal‐hearing young adults. The detection thresholds for a 20 ms increment added in the middle of a 420 ms pedestal of either 500 or 4000 Hz were obtained as a function of pedestal level. On‐frequency or off‐frequency maskers were either 30‐ or 150‐Hz wide random‐phase Gaussian noise (RPN) or low‐noise noise (LNN), centered either at 500 or 4000 Hz. The envelope fluctuation in LNN is minimal compared to that in RPN. The detection thresholds were greater in RPN than LNN in on‐frequency, but not in off‐frequency masking conditions. The detection thresholds increased at the wider masker bandwidth. Results suggest that listeners may rely on a decision process that is bas...

Effects of speakers' language background on speech perception in adults

With an increasingly changing sociolinguistic environment in the U.S., clinicians are challenged to accurately evaluate individuals’ speech perception under realistic everyday listening conditions. This study investigated how contextual linguistic information, speakers’ language background, speech-rate, and background noise individually and interactively affect listeners’ ability to recognize sentences. Ten normal-hearing American English native speakers, aged 20–22 years were recruited as listeners. Eight normal-hearing native speakers of four different languages (American English, Chinese-Mandarin, German, and Spanish) were recruited as speakers. These speakers read 120 sentences, 60 high-predictability and 60 low-predictability sentences from the revised Speech in Noise test (Bradlow and Alexander, 2007), which served as test stimuli. The listeners reported the target word of each sentence presented either in quiet or in multitalker babble. Results revealed that noise, high speech-rate, and lack of con...

Effects of linguistic background and noise on perception of fricatives

This study examined how native American English (AE) speakers perceived fricatives spoken by native AE and Bangladeshi Bengali (BB) speakers in quiet and noise. Participants included seven normal-hearing adults between 20 to 26 years of age. Participants listened to speech tokens of five fricatives, /s/, /z/, /∫/, /f/, and /v/ in the initial, medial, and final positions in the context of the point vowels /a/, /u/, /i/. Multitalker babble (MTB), speech noise, and three narrow bands of noise, 1000–2000 Hz, 2000–4000 Hz, and 500–5000 Hz at 45 dB SPL, 65 dB SPL and 85 dB SPL were used. The results suggested that listeners perceived fricatives significantly better when spoken by AE compared to BB speakers, and in quiet than in noise, especially in MTB. Listeners had the most difficulty with /z/, followed by /s/, /v/, /f, and /∫/ respectively, when tokens were spoken by BB speakers. This study may have implications for accent reduction therapy as well as for teaching English to English-language learners, especi...

Masking of low-frequency signals by high-frequency, high-level narrow bands of noise.

Low-frequency masking by intense high-frequency noise bands, referred to as remote masking (RM), was the first evidence to challenge energy-detection models of signal detection. Its underlying mechanisms remain unknown. RM was measured in five normal-hearing young-adults at 250, 350, 500, and 700 Hz using equal-power, spectrally matched random-phase noise (RPN) and low-noise noise (LNN) narrowband maskers. RM was also measured using equal-power, two-tone complex (TC2) and eight-tone complex (TC8). Maskers were centered at 3000 Hz with one or two equivalent rectangular bandwidths (ERBs). Masker levels varied from 80 to 95 dB sound pressure level in 5 dB steps. LNN produced negligible masking for all conditions. An increase in bandwidth in RPN yielded greater masking over a wider frequency region. Masking for TC2 was limited to 350 and 700 Hz for one ERB but shifted to only 700 Hz for two ERBs. A spread of masking to 500 and 700 Hz was observed for TC8 when the bandwidth was increased from one to two ERBs. Results suggest that high-frequency noise bands at high levels could generate significant low-frequency masking. It is possible that listeners experience significant RM due to the amplification of various competing noises that might have significant implications for speech perception in noise.

Off-frequency masking effects on intensity discrimination

Intensity discrimination, where a listener detects an intensity increment in an equal duration sinusoid or pedestal, is often used as a measure of intensity resolution. Intensity discrimination may be considered as tone-in-tone masking, where the pedestal is the masker and the increment is the signal. Despite the similarity between intensity discrimination and tone-in-noise masking, research suggests that a high-pass noise outside the critical band centered on the signal frequency adversely affects listeners’ intensity-discrimination thresholds. The present study examines the limits of off-frequency masking effects on intensity discrimination in five normal-hearing young adults. Detection thresholds for a 50-ms increment, added to a 50-ms-long 1000-Hz pedestal in phase, were obtained in quiet and notched-noise (NN) conditions. The pedestal and noise levels were 60 dB SPL. NN stimuli were generated by filtering telegraph noise. The low-frequency cutoffs of the NN-notches were 188, 250, 375, 500, and 750 Hz...

Relative contributions of on- and off-frequency spectral cues in increment detection

Increment detection is often used as a measure of intensity resolution, where a listener detects a brief increment in intensity in a longer duration stimulus. Recent findings suggest that listeners rely on multiple cues rather than on-frequency energy cues. The relative contributions of on- and off-frequency cues are evaluated in five normal-hearing young adults. Detection thresholds for a 50-ms increment added in the middle of a 450-ms, 1000-Hz pedestal were obtained in quiet, telegraph noise (TN), and notched-noise (NN) conditions. The pedestal level was 60 dB SPL. NN stimuli were generated by filtering TN while keeping the notch geometrically centered at 1000 Hz. The NN-notches were also varied by increasing the notch-width unevenly on either side of the center frequency. The low-frequency cutoffs of the NN-notches were 125, 250, 500, and 750 Hz while the high-frequency cutoffs were 1500, 2000, and 4000 Hz. Although the NN-notches were centered at 1000 Hz, increment detection thresholds were poorer in ...

A hybrid procedure for psychometric function estimation.

Threshold, defined as the stimulus level required for a predefined percent correct response, is often used to measure a listener’s performance. Psychometric functions (PFs) provide better insight to the underlying decision process. Two parameters, threshold (α) and slope (β), are sufficient to define a PF. Estimation of α and β from individual trials of adaptive procedure tracks is time‐consuming, costly, and dependent upon the step‐size choice. A procedure has been developed and written in MATLAB, which provides stable PF parameter estimates of α and β and also their confidence intervals. The procedure consists of three stages. First, signals are presented adaptively to estimate levels corresponding to 71% and 87% correct. Second, the signal is pseudo‐randomly presented at one of five fixed levels equally spaced over that range. Finally, the program implements a maximum‐likelihood procedure updated after every trial to estimate signal levels corresponding to 63%, 71%, 76%, 79%, and 87% correct. After 140...