Increase Speech Intelligibility Research Articles

Combining speech enhancement and auditory feature extraction for robust speech recognition

A major deficiency in state-of-the-art automatic speech recognition (ASR) systems is the lack of robustness in additive and convolutional noise. The model of auditory perception (PEMO), developed by Dau et al. (T. Dau, D. Püschel, A. Kohlrausch, J. Acoust. Soc. Am. 99 (6) (1996) 3615–3622) for psychoacoustical purposes, partly overcomes these difficulties when used as a front end for automatic speech recognition. To further improve the performance of this auditory-based recognition system in background noise, different speech enhancement methods were examined, which have been evaluated in earlier studies as components of digital hearing aids. Monaural noise reduction, as proposed by Ephraim and Malah (Y. Ephraim, D. Malah, IEEE Trans. Acoust. Speech Signal Process. ASSP-32 (6) (1984) 1109–1121) was compared to a binaural filter and dereverberation algorithm after Wittkop et al. (T. Wittkop, S. Albani, V. Hohmann, J. Peissig, W. Woods, B. Kollmeier, Acustica United with Acta Acustica 83 (4) (1997) 684–699). Both noise reduction algorithms yield improvements in recognition performance equivalent to up to 10 dB SNR in non-reverberant conditions for all types of noise, while the performance in clean speech is not significantly affected. Even in real-world reverberant conditions the speech enhancement schemes lead to improvements in recognition performance comparable to an SNR gain of up to 5 dB. This effect exceeds the expectations as earlier studies found no increase in speech intelligibility for hearing-impaired human subjects.

Auditory vs visual speech timing cues as external rate control to enhance verbal intelligibility in mixed spastic ataxic dysarthric speakers: a pilot study

Metronome, singing, and board pacing were used as external rate control techniques for the purpose of comparing the effectiveness of auditory and visual speech timing cues for reducing speech rate and increasing intelligibility in three traumatically brain injured mixed spastic ataxic dysarthric speakers. A single system design with baseline reversal (ABACAD) was used in this preliminary investigation. Results demonstrated statistically significant (p < 0.05) changes in increased speech intelligibility during all three pacing conditions for the two more involved subjects. Differences between treatment conditions were not statistically significant. However, auditory metronome cuing showed the best results for the two subjects who benefited from rate control. Lower baseline intelligibility was strongly correlated with higher benefit from rate control. Furthermore, the two auditory rhythmic pacing conditions exhibited a close synthronization effect between the frequency rate of the cue and speech rate. Significant correlation coefficients between decreased speech rate and increased intelligibility were only found for the two more involved subjects. These findings suggested a differential benefit of slowing speech rate to improve intelligibility contingent upon severity of speech deficits.

Increasing speech intelligibility in children with autism.

Accumulating studies are documenting specific motivational variables that, when combined into a naturalistic teaching paradigm, reliably influence the effectiveness of language teaching interactions for children with autism. However, the effectiveness of this approach has not yet been assessed with respect to improving speech intelligibility. The purpose of this study was to systematically compare two intervention conditions, a Naturalistic approach (which incorporated motivational variables) vs. an Analog (more traditional, structured) approach, with developmentally similar speech sounds equated within and across conditions for each child. Data indicate that although both methods effectively increased correct production of the target sounds under some conditions, functional use of the target sounds in conversation occurred only when the naturalistic procedures were used during intervention. Results are discussed in terms of pivotal variables that may produce improvements in speech sounds during conversational speech.

Effects of Temporal Alterations on Speech Intelligibility in Parkinsonian Dysarthria

The effect of two types of temporal alterations, paced and synthetic, on the intelligibility of parkinsonian dysarthric speech was investigated. Six speakers with idiopathic Parkinson's disease served as subjects. Paced temporal alterations were created by slowing each speaker to 60% of his/her habitual speaking rate. The synthetic alterations were created by modifying the habitual rate speech samples using digital signal processing. Three types of synthetic alterations were examined: Pause Altered, Speech Duration Altered, and Pause and Speech Duration Altered. The 60% of habitual speaking rate condition was more intelligible than the synthetic conditions. In addition, none of the synthetic alterations were found to be more intelligible than samples produced at habitual speaking rates. The results suggest that simple alterations of speech signals do not explain the differences in intelligibility that have been observed when parkinsonian dysarthric speakers reduce speaking rates. Reasons for the failure of synthetic alterations to increase speech intelligibility scores are discussed.

Speech enhancement based on physiological and psychoacoustical models of modulation perception and binaural interaction.

A novel approach for analyzing and filtering speech is described and evaluated which utilizes the "modulation spectrogram," i.e., the two-dimensional representation of modulation frequencies versus center frequency as a function of time. This approach is based on physiological findings of a tonotopical organization of modulation frequencies perpendicular to carrier frequencies as well as psychoacoustical findings of "modulation tuning curves." In addition, an interaction is assumed between the representation of modulation frequencies and the representation of auditory space as described by physiological and psychological models of binaural hearing. A noise-reduction algorithm based on this approach was implemented and tested which enhances or suppresses each combination of modulation frequency and center frequency according to its phase and intensity relation between the two input signals (i.e., both stereo channels of a dummy-head recording). When tested in several situations with interfering speakers and background noise both in anechoic and reverberant environment, the algorithm provided a small but a very robust increase in speech intelligibility which corresponds to approximately 2 dB in signal-to-noise ratio. Possible applications of this algorithm are noise reduction in adverse acoustical situations, digital hearing aids, processing schemes and preprocessing for speech recognition.

Voice rehabilitation after laryngectomy with voice prostheses. Results of a prospective follow-up study

Although the results of surgical rehabilitation by means of voice prostheses are on the average better than rehabilitation via oesophageal speech, the tracheoesophageal puncture (TEP)-technique has so far not been widely used in Germany. The majority of hospitals still prefer the "traditional" method of voice rehabilitation using oesophageal speech. The present prospective study was undertaken to compare the results of postlaryngectomy vocal rehabilitation, if patients were offered the surgical voice rehabilitation via voice prosthesis as an alternative to oesophageal speech. Taking into account all the patients who underwent laryngectomy from 1989 until 1990 in Tübingen, primary surgical voice rehabilitation was performed in 44 out of 54 patients (81.5%). Interestingly enough, 34 patients who underwent laryngectomy were able to perform communication via the telephone on the day of their discharge. Moreover, one-third of the laryngectomised patients showed a significant increase in speech intelligibility within the first six months after laryngectomy. 36 patients with laryngectomy were able to attain proficiency 6 months after surgery. In 12 patients the prosthesis had to be removed, since either phonation was impossible or patients successfully learned and preferred oesophageal speech. In conclusion, independent of the method of voice rehabilitation (prosthesis, electrolarynx, oesophageal speech), our results support the hypothesis that a voice rehabilitation regimen will yield a higher rehabilitation rate of patients if rehabilitation via surgical voice is offered as an alternative to learning the oesophageal voice. Therefore, it seems to be advisable that patients are allowed to have the choice between surgical rehabilitation and oesophageal speech restoration.(ABSTRACT TRUNCATED AT 250 WORDS)

The use of strategies to increase speech intelligibility in cerebral palsy: An experimental evaluation

The effects of strategies on the intelligibility of cerebral-palsied speech were experimentally evaluated. Listeners were asked to identify words in sentences spoken by subjects whose dysarthria was rated to be either moderate or severe. Results indicated that strategies increased intelligibility and that different strategies were appropriate at different levels of severity. Use of initial letter cueing achieved the greatest gains for the more severely impaired speakers but, for the less severely impaired, the gains achieved by providing visual cues were the same as those achieved by letter cueing and could not be improved upon by repetition. Listeners' experience of dysarthric speech and the redundancy of the test sentences were also manipulated. Previous experience did not influence recognition and redundancy only appeared to have an influence at higher levels of intelligibility. The role of strategies in improving intelligibility is discussed. The results suggest that many speakers will be unable to use simple strategies to improve intelligibility to levels that may be needed for adequate information transfer.

Is AGC beneficial in hearing aids?

Three different functions of Automatic Gain Control (AGC) circuits in hearing aids are distinguished and the evidence for their benefits is considered. The value of AGC's function as a relatively distortion-free means of limiting output has been well established. With regard to compression, the benefit of short-term or 'syllabic' compression has not been demonstrated convincingly. Most evaluations of this type of AGC have looked for increase in speech intelligibility, but theoretical predictions of its effect do not appear to take account of the acoustic cues to consonant contrasts actually used by hearing impaired people, and empirical studies have often used listening conditions which do not give a realistic test of benefit. Relatively little attention has been paid to long-term compression, or to the effect of AGC on comfort rather than intelligibility. Listening tests carried out at the RNID and reported here have shown that AGC can benefit hearing aid users by allowing them to listen to a wider range of sound levels without either strain or discomfort, and, if time constants are well chosen, without adverse effects on speech intelligibility in quiet or in noise.

The effect of compression limiting on the speech intelligibility of sensorineural hearing impaired listeners

The purpose of this study was the evaluation of multiband compression limiting designed to protect hearing-impaired listeners from high-level speech and environmental sounds. The relative performance of two linear and four compression-limiting systems has been measured for two subjects with flat sensorineural losses. The test materials, consisting of nonsense CVC syllables spoken by male and female speakers, were presented over a 30-dB range of input levels. The subjects chose an overall listening level consistent with both maximum intelligibility over the range of input levels and long-term listening comfort. Preliminary results indicate that both higher average presentation levels and increased speech intelligibility scores are associated with compression limiting relative to linear amplification. [Work supported by NIH.]

Evaluation of a correlation subtraction method for enhancing speech degraded by additive white noise

An intelligibility test was performed to evaluate a correlation subtraction method for enhancement of degraded speech due to additive white noise. Results indicate that such a scheme does not significantly increase speech intelligibility at the S/N ratios where the intelligibility scores of unprocessed speech range between 20 and 70 percent.

Evaluation of an adaptive comb filtering method for enhancing speech degraded by white noise addition

An intelligibility test was performed to evaluate an adaptive comb filtering method proposed by Frazier [2] for enhancement of degraded speech due to additive white noise. Results indicate that independent of S/N ratio the adaptive comb filtering scheme does not increase speech intelligibility.

Vibrotactile encoding of speech signals

Preliminary results of a research effort concerned with the sensory encoding of speech by humans through the vibrotactile system are described. The research involves the use of a hardware system which includes a frequency‐transformation device [see Finkenzeller, P., Habilitation Thesis, Universität Erlangen, Erlangen, Germany (1973)] and a vibrotactile transducer which is used to present the transformed speech signal to the hand or arm. The transformation device operates on the speech signal essentially in real time and is capable of lowering the spectral content of the waveform while preserving part of the information available in the fine structure. In this paper, the transformation device and results of early speech‐discrimination experiments are described. These initial experiments have primarily involved the psychophysical measurement (using an A‐B‐X paradigm) of intelligibility of words selected from the W‐22 and Modified Rhyme Test lists. The results suggest that the utilization of appropriate spectral and intensive transformations of the speech signal lead to increased speech intelligibility when speech is presented via the vibrotactile system.

Peculiarities of Wiener Class Systems and their Exploitation for Speech Signal Prediction

In this paper Wiener class system consisting of infinite impulse response dynamic subsystem and sigmoid nonlinearity is employed for speech modelling. The experiments on Lithuanian vowel "ū" with different initial phase of the signal are done. The average of mean square errors of prediction with the Wiener system model are smaller approximately by 18 % for the training dataset, 17 % for the validation dataset and 14 % for the testing dataset in comparison to Linear Prediction Coding model. Additional experiments with 6 different Lithuanian vowels on average showed more than 13 % improvement in MSE of prediction with the Wiener system model. Moreover, the use of Wiener system instead of IIR model showed improvement in MSE of long-term prediction for five Lithuanian vowels on average by more than 5 %. Application of results in telecommunication systems could possibly decrease data transfer rates, increase speech intelligibility and naturalness of speech synthesis. Ill. 5, bibl. 8, tabl. 2 (in English; abstracts in English and Lithuanian).http://dx.doi.org/10.5755/j01.eee.111.5.368