Effects of spectral smearing on sentence recognition by adults and children
Joanna H. Lowenstein, Eric Tarr, Susan Nittrouer
Abstract
Joanna H. Lowenstein, Eric Tarr, Susan Nittrouer
Abstract
Children’s speech perception depends on dynamic formant patterns more than that of adults. Spectral smearing of formants, as found with the broadened auditory filters associated with hearing loss, should disproportionately affect children because of this greater dependence on formant patterns. Making formants more prominent, on the other hand, may result in improved recognition. Adults (40) and children age 5 and 7 (20 of each age) listened to 75 four-word syntactically correct, semantically anomalous sentences processed so that excursions around the mean spectral slope were sharpened by 50% (making individual formants more prominent), flattened by 50% (smearing individual formants), or left unchanged. These sentences were presented to children and to half of the adults in speech-shaped noise at 0 dB SNR. The rest of the adults listened to the sentences at -3 dB SNR. Results indicate that all listeners did more poorly with the smeared formants, with 5-year-olds showing the largest decrement in performance at 0 dB SNR. However, adults at -3 dB SNR showed an even greater decrement in performance. Making formants more prominent did not improve recognition, perhaps due to harmonic-formant mismatches. Thus, there is reason to explore processing strategies that might enhance formant prominence for listeners with hearing loss.
A significance statement is not available in the OpenAlex record.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
Children’s speech perception depends on dynamic formant patterns more than that of adults. Spectral smearing of formants, as found with the broadened auditory filters associated with hearing loss, should disproportionately affect children because of this greater dependence on formant patterns. Making formants more prominent, on the other hand, may result in improved recognition. Adults (40) and children age 5 and 7 (20 of each age) listened to 75 four-word syntactically correct, semantically anomalous sentences processed so that excursions around the mean spectral slope were sharpened by 50% (making individual formants more prominent), flattened by 50% (smearing individual formants), or left unchanged. These sentences were presented to children and to half of the adults in speech-shaped noise at 0 dB SNR. The rest of the adults listened to the sentences at -3 dB SNR. Results indicate that all listeners did more poorly with the smeared formants, with 5-year-olds showing the largest decrement in performance at 0 dB SNR. However, adults at -3 dB SNR showed an even greater decrement in performance. Making formants more prominent did not improve recognition, perhaps due to harmonic-formant mismatches. Thus, there is reason to explore processing strategies that might enhance formant prominence for listeners with hearing loss.
Key concepts: Formant, Audiology, Sentence, Noise (video), Psychology, Acoustics, Speech recognition, Mathematics