Recognition Of Stop Consonants Research Articles

Comparing the Recognition of Stop Consonants in Persian Between Adults Aged 30-39 and 40-49 Years

Background and Objectives Due to the improvement of living conditions in recent decades, there is an important issue called quality of life (QoL). One of the important factors that affect the QoL is the ability to communicate with others. Everyday communication is usually interfered by noise. Consonants play an important role in understanding the meaning of words. Since stop consonants are dependent on the following vowels, they are more vulnerable to the masking effects of noise. The present study aims to evaluate the effect of age, signal-to-noise ratio (SNR), and concurrent vowels on the recognition of stop consonants in Persian. Subjects and Methods The present cross-sectional study was conducted on 60 adults with normal hearing in two age groups of 30-39 and 40-49 years. After the auditory and speech assessments, the recognition of stop consonants in the form of consonant-vowel-consonant syllables and in the presence of babble noise at three SNRs was compared between two age groups. Results The difference in the recognition score of stop consonants was significant between two age groups at the SNRs of 0, -5, and -10. There was also a significant difference between the two age groups regarding the recognition of stop consonants in the presence of some vowels at three SNRs. Conclusion With increasing age, the recognition ability of stop consonants in the presence of babble noises decreases. Increasing noise also reduces the recognition of stop consonants, especially those with some vowels.

Consonant Recognition Using Coarticulatory Cues in Individuals with Normal Hearing and Sensorineural Hearing Loss

Background and Objectives: The study investigated the role of coarticulatory cues in the perception of consonants in Malayalam and its temporal window. It also compared normal hearing individuals and individuals with sensorineural hearing loss (SNHL) for their ability to utilize coarticulatory cues for the perception of consonants. Methods: The study used a quasi- experimental post-test only mixed research design. Fifteen normal-hearing individuals and 15 individuals with SNHL who were native speakers of Malayalam participated in the study. The stimuli included consonant-vowel syllables in their original and truncated forms. The forward-gating method was used to generate the truncated tokens. The participants were assessed for their consonant recognition in closed-set conditions. Results: There was a significant difference in the temporal window of the utility of coarticulatory cues across consonants and also between the two groups of participants. Conclusions: In normal-hearing individuals, coarticulatory cues are useful for the recognition of stop consonants, nasals as well as fricatives, with the maximum temporal window of utility seen in nasals. However, individuals with SNHL fail to utilize the available coarticulatory cues to recognize the consonants.

Burst Onset Landmark Detection and Its Application to Speech Recognition

The reliable detection of salient acoustic-phonetic cues in speech signal plays an important role in speech recognition based on speech landmarks. Once speech landmarks are located, not only can phone recognition be performed, but other useful information can also be derived. This paper focuses on the detection of burst onset landmarks, which are crucial to the recognition of stop and affricate consonants. The proposed detector is purely based on a random forest technique, which belongs to an ensemble of tree-structured classifiers. By adopting a special asymmetric bootstrapping method, a series of experiments conducted on the TIMIT database demonstrate that the proposed detector is an efficient and accurate method for detecting burst onsets. When the detection results are appended to mel frequency cepstral coefficient vectors, the augmented feature vectors enhance the recognition correctness of hidden Markov models in recognizing stop and affricate consonants in continuous speech.

Multichannel Compression Hearing Aids: Perceptual Considerations

Multichannel Compression Hearing Aids: Perceptual Considerations

The representation of noise vocoded speech in the auditory nerve of the chinchilla: Physiological correlates of the perception of spectrally reduced speech

This study investigated the neural representation of naturally produced and noise vocoded speech signals in the auditory nerve of the chinchilla. The syllables / b ɔ / , / d ə ̢ / , / t u / and / p ə ̢ / produced by male speakers were used to synthesize noise vocoded speech stimuli containing one, two, three and four bands of envelope modulated noise. The ensemble response of the auditory nerve, computed by pooling the PST histograms across many auditory nerve fibers, revealed temporal patterns in the responses to the natural tokens that uniquely identified the stop consonants. The responses to the 3- and 4-band noise vocoded tokens contained temporal patterns that were nearly identical to those observed for the natural tokens, while the responses to the 1- and 2-band tokens were significantly different ( p < 0.0001). The ALSR, ALIR and autocorrelation of the pooled PST histograms represented the detail of the frequency spectrum for a naturally produced vowel, while the driven rate was unreliable. Each of these spectral analyses failed to reveal significant information about the noise vocoded vowels. These results suggest that temporal patterns in the responses of the auditory nerve can provide the cues necessary for the recognition of noise vocoded stop consonants.

The influence of noise on vowel and consonant cues

This study assessed the acoustic and perceptual effect of noise on vowel and stop-consonant spectra. Multi-talker babble and speech-shaped noise were added to vowel and stop stimuli at -5 to +10 dB S/N, and the effect of noise was quantified in terms of (a) spectral envelope differences between the noisy and clean spectra in three frequency bands, (b) presence of reliable F1 and F2 information in noise, and (c) changes in burst frequency and slope. Acoustic analysis indicated that F1 was detected more reliably than F2 and the largest spectral envelope differences between the noisy and clean vowel spectra occurred in the mid-frequency band. This finding suggests that in extremely noisy conditions listeners must be relying on relatively accurate F1 frequency information along with partial F2 information to identify vowels. Stop consonant recognition remained high even at -5 dB despite the disruption of burst cues due to additive noise, suggesting that listeners must be relying on other cues, perhaps formant transitions, to identify stops.

Voiced–unvoiced classification for recognition of stop consonants

Significant error in stop consonant recognition is caused by the confusion between voiced stop consonants and their unvoiced counterparts. The recognition is based on HMM’s which use 12 MFCC’s and energy with their time derivatives. The voicing state is the distinctive feature of homorganic stop consonants. According to recognition error-rate analysis, it seems that the cepstral feature does not accurately represent the voicing state of the modeled phone. For this purpose, a voiced–unvoiced classifier in conjunction with HMM’s is proposed to improve the recognition of stop consonants. The recognition is done in two passes. In the first pass, a phone recognizer uses well-trained HMM’s to identify a stop consonant. This pass provides the recognized stop consonant in addition to the log probability. In the second pass, the voiced–unvoiced classifier checks if the voicing state of the phone segment matches with its phonetic description. In the case of mismatch and low probability of recognition, the voiced (unvoiced) consonant is swapped with its corresponding unvoiced (voiced) counterpart. Recognition results are presented in terms of error rate, using different techniques of voiced–unvoiced classification. This method reduces the recognition error rate of stop consonants. [Research supported by DARPA Contract Nos. DABT63-93-C-0037, N66001-96-C-8510, and NSF Contract No. IRI-9618854.]

Acoustic-phonetic features for the automatic recognition of stop consonants

Despite the recent successes in the field of automatic speech recognition, more research is still needed in order to understand the variability of speech and the acoustic characteristics of speech sounds in different contexts and for different speakers. In this paper, the acoustic-phonetic characteristics and the automatic recognition of the American English stop consonants are investigated. The acoustic features that exist in the literature are evaluated and new features are proposed. To test the value of the extracted features, a knowledge-based acoustic-phonetic system for the automatic recognition of stops, in speaker independent continuous speech, is proposed. The system uses an auditory-based front-end processing and incorporates new algorithms for the extraction and manipulation of the acoustic-phonetic features that proved to be rich in their information content. Several features, which describe the burst frequency, formant transitions, relative amplitude, spectral shape, and duration, are combined in the recognition process. Recognition accuracy of 95% for voicing detection and 90% for place of articulation detection are obtained for TIMIT database continuous speech of multiple speakers from different dialect regions. The obtained results are analyzed and compared to previous work. [Work was supported by Catalyst Foundation.]

Detection and recognition of stop consonants by normal-hearing and hearing-impaired listeners.

This study examined the possibility that hearing-impaired listeners, in addition to displaying poorer-than-normal recognition of speech presented in background noise, require a larger signal-to-noise ratio for the detection of the speech sounds. Psychometric functions for the detection and recognition of stop consonants were obtained from both normal-hearing and hearing-impaired listeners. Expressing the speech levels in terms of their short-term spectra, the detection of consonants for both subject groups occurred at the same signal-to-noise ratio. In contrast, the hearing-impaired listeners displayed poorer recognition performance than the normal-hearing listeners. These results imply that the higher signal-to-noise ratios required for a given level of recognition by some subjects with hearing loss are not due in part to a deficit in detection of the signals in the masking noise, but rather are due exclusively to a deficit in recognition.

A study on the automatic recognition of voiceless unaspirated stops

This paper presents a method of feature extraction for the automatic recognition of voiceless unaspirated stop consonants in Mandarin speech. The features are derived from the spectrographic acoustic patterns of syllable-initial voiceless unaspirated stops /p,t,k/, which include the burst spectrum, the formant transition, and the voice onset time. A normalization process for the second and the third formants at the voice onset is proposed. Based on these derived features, Bayes classifiers and a layered neural net are applied to classify the places of articulation of these stop consonants. The experiments show that the derived features are robust and efficient for speaker-independent speech recognition, and the neural net is a preferable choice in the classification of these stops in multiple contexts.

Associations among Frequency and Temporal Resolution and Consonant Recognition for Hearing-impaired Listeners.

To examine the association between frequency and temporal resolution and speech recognition, auditory filter parameters, rates of masking decay, and stop-consonant recognition were determined for 9 normal-hearing and 24 hearing-impaired subjects. Speech-presentation levels were selected for each subject based on articulation index (AI) predictions. Results suggest that auditory filter widths and dynamic ranges were strongly correlated with pure-tone threshold, while time constants were not. Stop-consonant recognition scores for most hearing-impaired listeners were not significantly poorer than predicted by the AI model. Furthermore, differences between observed recognition scores and those predicted by the AI were poorly correlated with derived psychophysical parameter values, suggesting that measures of frequency and temporal resolution and speech recognition may appear to be associated primarily because of their dependence on auditory threshold.

Spectral shape factors for speaker‐independent automatic recognition of stop consonants

A series of automatic recognition experiments was conducted with naturally produced English stop consonants /b,p,d,t,g,k/ in syllable initial position. The objectives of the experiments were to investigate in detail the effectiveness of spectral shape factors, as both dynamic and static features for automatic recognition of stop consonants. Spectral shape factors were computed as the discrete cosine transform coefficients (DCTCs) of the magnitude spectra. The database used in these experiments consisted of 2481 CVC syllables spoken in isolation by ten males, ten females, and ten children. In all experiments, 15 speakers were used to train the classifier and the other 15 speakers were used for evaluation. For the case of dynamic features, DCTCs were computed over a 50‐ms interval beginning with the burst using 7‐ms frames spaced every 5 ms. For the static case, the DCTCs were computed from one 25.6‐ms frame beginning at the burst. Automatic classification results, based on the test data, were 87.2% for the dynamic spectral shape features versus 60.1% for the static case. Dynamic features timed to begin with the formant transition area resulted in 44.6% recognition rates. Control experiments with formants resulted in much lower recognition rates under every condition tested. In summary, these results are in agreement with the theory that dynamic spectral shape, spanning an interval of approximately 50 ms beginning at the burst, carries most of the information for initial stop consonants [D. Kewley‐Port, J. Acoust. Soc. Am. 73, 322–335 (1983)]. [Work supported by NSF.]

Auditory filter characteristics and consonant recognition for hearing-impaired listeners

To examine the association between frequency resolution and speech recognition, auditory filter parameters and stop-consonant recognition were determined for 9 normal-hearing and 24 hearing-impaired subjects. In an earlier investigation, the relationship between stop-consonant recognition and the articulation index (AI) had been established on normal-hearing listeners. Based on AI predictions, speech-presentation levels for each subject in this experiment were selected to obtain a wide range of recognition scores. This strategy provides a method of interpreting speech-recognition performance among listeners who vary in magnitude and configuration of hearing loss by assuming that conditions which yield equal audible spectra will result in equivalent performance. It was reasoned that an association between frequency resolution and consonant recognition may be more appropriately estimated if hearing-impaired listeners' performance was measured under conditions that assured equivalent audibility of the speech stimuli. Derived auditory filter parameters indicated that filter widths and dynamic ranges were strongly associated with threshold. Stop-consonant recognition scores for most hearing-impaired listeners were not significantly poorer than predicted by the AI model. Furthermore, differences between observed recognition scores and those predicted by the AI were not associated with auditory filter characteristics, suggesting that frequency resolution and speech recognition may appear to be associated primarily because both are degraded by threshold elevation.

Speech cue enhancement in intervocalic stops

The place of articulation feature for stop consonants is subject to many errors in speech processing by hearing-impaired listeners. Attempts to improve the recognition of initial and final stop consonants by lowering the level of the first formant or-with a different approach-by narrowing the formant bandwidth of the first five formants only very partially led to satisfactory results. Intervocalic stops were used in the present investigation, on the one hand because the spectral information is represented twice (in the VC as well as in the CV-transition) and, on the other hand, because the closure duration offers additional information to the listener. The modification of bandwidth led to no noticeable improvement in the /b, d, g/ discrimination. A change of the closure durations affected the identification of the stops, especially the /b/. The modification of the transitions of the second and third formants optimized the recognition rates for /b/ and /g/.

標準パターンの時間軸掃引による音素区間からの破裂子音,鼻子音の認識

標準パターンの時間軸掃引による音素区間からの破裂子音,鼻子音の認識

Effects of hearing loss on utilization of short-duration spectral cues in stop consonant recognition.

The purpose of this experiment was to evaluate the utilization of short-term spectral cues for recognition of initial plosive consonants (/b,d,g/) by normal-hearing and by hearing-impaired listeners differing in audiometric configuration. Recognition scores were obtained for these consonants paired with three vowels (/a,i,u/) while systematically reducing the duration (300 to 10 ms) of the synthetic consonant-vowel syllables. Results from 10 normal-hearing and 15 hearing-impaired listeners suggest that audiometric configuration interacts in a complex manner with the identification of short-duration stimuli. For consonants paired with the vowels /a/ and /u/, performance deteriorated as the slope of the audiometric configuration increased. The one exception to this result was a subject who had significantly elevated pure-tone thresholds relative to the other hearing-impaired subjects. Despite the changes in the shape of the onset spectral cues imposed by hearing loss, with increasing duration, consonant recognition in the /a/ and /u/ context for most hearing-impaired subjects eventually approached that of the normal-hearing listeners. In contrast, scores for consonants paired with /i/ were poor for a majority of hearing-impaired listeners for stimuli of all durations.

Stop consonant recognition and audibility in normal and hearing‐loss subjects

This study examined the effect of spectral cue audibility on the discrimination of CV stop consonants in normal‐hearing and hearing‐impaired adults. Six synthetic speech tokens, each differing only in the initial 40‐ms consonant portion, were presented to subjects in randomized lists. Performance‐intensity functions and relative information transmitted were calculated from the results of a six‐alternative, closed‐set response task, In both normal‐hearing and hearing‐impaired subjects, recognition performance as a function of level differed among the six consonants. The first 40 ms of each CV were analyzed via FFT using an exponential window. The resulting spectral array was passed through a sliding‐filter model of the auditory system to account for the proportional bandwidth filtering properties of the ear. This allowed the spectral data to be displayed in comparison to a subject's pure‐tone thresholds. Differences in the amount of predicted audible spectral speech cues resulted in observable differences in the recognition performance of individual consonants.

Recognition of stop consonants in Japanese words using local spectral peaks.

This paper describes the experiments on recognition of stop consonants in continuous speech by use of local spectral peaks. The spectrum obtained through a band pass filter bank is sampled every 10 ms. The sampled spectrum is represented by a binary valued vector in which every element denotes the presence or absence of a local peak. The frequency distribution of local spectral peaks in 40 ms from the burst frame is transformed into the feature vector. And the conditional probability of the feature vector is used for the recognition. The experiments were carried out using 212 Japanese words uttered by 10 males and 10 females. The stop consonants are discriminated at a comparatively high rate using only the local spectral peaks. To improve the recognition rates, the differences between successive spectra are used. The recognition rates for the unvoiced stops for the 10 male and 10 female speakers are 82.6% and 79.1 % respectively, by the leaving-one-out experiments ; those for the voiced stops are 74.5 % and 66.1 %. It turns out that the local spectral peaks and the temporal changes in the spectra are very significant features for discriminating stop consonants.

Factors underlying errors in the recognition of stop consonants by listeners with cochlear hearing impairment

In a series of experiments we have assessed the contribution of selected stimulus factors to the recognition of stop consonants by listeners with mild to severe hearing impairments. Errors in recognition do not appear to be conditioned in a significant manner by a high amplitude first formant or by brief duration formant transitions. Errors are conditioned by the spectral tilt at signal onset and by the duration of the steady‐state spectrum at signal onset. In severely hearing‐impaired listeners, discrimination among the stops appears dependent on the presence of distinctive temporal properties among the signals. In general, our results point to the critical role played by accurate temporal resolution in maintaining accurate speech intelligibility.

Knowledge‐based recognition of voiceless stop consonants

Recent experiments [M. Bush et al., Proc. 1983 IEEE ICASSP] have demonstrated the ability of a trained spectrogram reader to identify initial stops in /CVb/ syllables from a table of numerical acoustic measurements with approximately 80% accuracy. This paper discusses an automatic system for discriminating between the voiceless plosives (/p,t,k/)which is based on the features and rules identified in these experiments. Ten binary features are extracted from two linear prediction spectra which are computed during the 35 ms following the consonant release. Typical features include “back‐k‐release‐spectrum” and “compact‐release‐spectrum.” The features are detected by examining the frequencies and amplitudes of the local maxima and minima of the two LPC spectra, in a manner motivted by the actions of the human spectrogram reader. A simple statistical classifier is used to combine the outputs of the ten feature detectors. The classifier was trained on the 108 /p,t,k/ tokens of the multi‐speaker corpus used in the spectrogram reading experiment. When tested on the training data, the system achieved 96% correct recognition. When tested on two additional data sets of similar composition, the system achieved scores of 94% and 92%, respectively.