Candace Kamm

Speech recognition and the Articulation Index for normal and hearing‐ impaired listeners

The purpose of this experiment was to determine the applicability of the Articulation Index (AI) model for characterizing the speech recognition performance of listeners with mild-to-moderate hearing loss. Performance-intensity functions were obtained from five normal-hearing listeners and 11 hearing-impaired listeners using a closed-set nonsense syllable test for two frequency responses (uniform and high-frequency emphasis). For each listener, the fitting constant Q of the nonlinear transfer function relating AI and speech recognition was estimated. Results indicated that the function mapping AI onto performance was approximately the same for normal and hearing-impaired listeners with mild-to-moderate hearing loss and high speech recognition scores. For a hearing-impaired listener with poor speech recognition ability, the AI procedure was a poor predictor of performance. The AI procedure as presently used is inadequate for predicting performance of individuals with reduced speech recognition ability and should be used conservatively in applications predicting optimal or acceptable frequency response characteristics for hearing-aid amplification systems.

Form equivalence of the speech perception in noise (SPIN) test

An evaluation was conducted of the equivalence of the ten forms of the Speech Perception in Noise (SPIN) Test with normal hearing subjects. Each subject was tested monaurally on all ten forms. Twenty-five subjects were presented the materials at 80 dB SPL at a signal-to-babble ratio of - 1 dB. An additional 25 subjects heard the materials at 30 dB SPL, at a signal-to-babble ratio of + 3 dB. The data were analyzed for equivalence using a parallel tests model. A detailed rationale for the application of this model to the equivalence of the SPIN test is presented. The results indicated that a subset of seven lists fits the equivalence model for PH, PL, and difference (PH - PL) scores. Additionally, analysis of list-pair data (combination of companion forms) suggested that all five list-pairs resulted in equivalent performance.

Using an adaptive network to recognize demisyllables in continuous speech

A nonlinear, multilayer associative network was trained on a speech recognition task using continuous speech. Naturally spoken 14‐syllable “sentences” from one talker were preprocessed to produce a 15‐band spectral representation incorporating several transformations introduced by the peripheral auditory system on acoustic signals. Input nodes to the network represented a 150‐ms window through which the spectral representation passed in 2‐ms steps. A single layer of 20 hidden nodes was used. Output nodes represented seven initial demisyllables whose target values were specified based on a human listeners identification of the sounds heard during the input segment. The network was trained to criterion using a variant of the back‐propagation learning algorithm [Rumelhart et al., Nature 323, 533–536 (1986); Landauer et al., Proc. Cog. Sci. Soc., 531–536 (1987)]. A minimum‐error‐rate figure of merit (derived from signal detection theory) was used to evaluate the effect of the size of the training corpus on t...

The influence of middle-ear muscle contraction on auditory threshold for selected pure tones.

The influence of middle-ear muscle (MEM) contraction on auditory threshold has been measured for pure tones of 0.25, 0.5, and 1.5 kHz. The reflex-activating signal was a 3-kHz pure tone. Signal paradigms were chosen to reduce or eliminate the effects of binaural loudness summation, contralateral direct masking, and contralateral remote and backward masking effects, and to maximize the influence of MEM contraction. Results indicate that under no condition was behavioral threshold affected by the MEM contraction induced using a pure-tone stimulus of 3 kHz, 105 dB SPL.

Phonetic dependencies in syllables with full and reduced vowels

Speech synthesis using demisyllable concatenation relies on the assumption that initial and final demisyllables are phonetically independent of each other, aside from the need for smoothing the boundary. The success of demisyllable concatenation reported by Lovins, Macchi, and Fujimura [J. Acoust. Soc. Am. Suppl. 1 65, S130 (1979)] suggests that any phonetic dependencies that exist are not perceptually salient, at least for phrase‐final syllables containing stressed vowels. To asess the independence assumption for naturally spoken English CVC syllables, the influences the initial consonants on syllable‐final events and of final consonants on syllable‐initial events were examined for long‐duration syllables containing full vowels and shorter‐duration syllables containing full and reduced vowels. Reduced syllables showed larger variation in formant frequencies at vowel offset across initial consonants than full‐vowel syllables, including those with comparable vowel duration. Similarly, reduced syllables showed larger variation in formant frequencies at vowel onset across final consonants than syllables containing long‐duration full vowels. To maintain the contextual variation in reduced vowels in demisyllable synthesis, it may be necessary to devise more complex smoothing rules or to include several versions of reduced‐vowel demisyllables in the inventory.

Predicting effects of spectral shaping on speech recognition of hearing‐impaired listeners

The effects of spectral modification on speech recognition were investigated for sensorineural listeners: One group with flat audiometric configuration and a second group with sharply‐sloping high‐frequency hearing loss. Three spectral shapes were tested: Uniform frequency gain, high‐pass filtering, and a response shaped relative to loudness discomfort levels. Performance‐intensity functions were measured at four levels (from 80–95 dB SPL) using the CUNY Nonsense Syllable Test (NST) and the Synthetic Sentence Identification task (SSI), both presented monaurally under earphones against a background of multitalker babble. No significant differences in performance on the NST were observed between the two subject groups across all spectral shapes and presentation levels. On the SSI, performance of subjects with flat audiometric configuration was highest using the uniform frequency response, while performance of listeners with sloping hearing loss was poorest for the uniform spectral shape. The recognition dat...