Search

Recently Deep Denoising Autoencoders (DDAE) have shown state-of-the-art performance on various machine learning tasks. In this paper, the authors extended this approach to whispered speech recognition which is one of the most challenging problems in Automatic Speech Recognition (ASR). Namely, due to the profound differences between acoustic characteristics of neutral and whispered speech, the performance of traditional ASR systems trained on neutral speech degrades significantly when whisper is applied. This mismatch between training and testing is successfully alleviated with the new proposed system based on deep learning, where DDAE is applied for generating whisper-robust cepstral features. This system was tested and compared in terms of word recognition accuracy with conventional Hidden Markov Model (HMM) speech recognizer in an isolated word recognition task with a real database of whispered speech (WhiSpe). Three types of cepstral coefficients were used in the experiments: MFCC (Mel-Frequency Cepstral Coefficients), TECC (Teager-Energy Cepstral Coefficients) and TEMFCC (Teager-based Mel-Frequency Cepstral Coefficients). The experimental results showed that the proposed system significantly improves whisper recognition accuracy and outperforms traditional HMM-MFCC baseline, resulting in an absolute 31% improvement of whisper recognition accuracy. The highest word recognition rate of 92.81% in whispered speech was achieved with TECC feature. [ABSTRACT FROM AUTHOR]

Abstract: The purpose of the present study was to investigate (a) the distortion in production of word-initial friction duration in fricative /ʃ/, and (b) the perceptual discrimination between typical (normal) and atypical (prolonged or lengthened) friction duration. In the first experiment 80 school aged children pronounced word /ʃuma/, 40 of them have had typical friction duration, 15 shortened and 25 lengthened duration of /ʃ/, and all of them did not have any distortion in segment /uma/. Measurement and analysis of duration of both segments /ʃ/ and /uma/ were performed after segmentation of speech stimuli /ʃuma/. In second experiment friction duration of /ʃ/ has been synthesized in 17-step continuum from 135ms to 311ms in equal steps of 11ms, friction duration of 223ms has accepted as average typical value, and segment /uma/ was unchanged. Six speech-language pathologists (SLPs) listened to the synthesized stimuli /ʃuma/ and performed perception task –distortion contrast in two sessions. They discriminated the shortened and lengthened duration in comparison to typical duration of fricative /ʃ/ and results were interpreted by identification functions. The results of the first experiment indicate that distortion in production of friction duration of /ʃ/ does not affects duration of the other segment /uma/ in word /ʃuma/, and that these two durations are non-correlated. In perception of distortion in friction duration SLPs show high inter-listeners agreement and their identification functions showed very good discrimination in typical/atypical friction duration. Learning outcomes: Readers will be able to: (1) understand the nature of distortion in friction duration in terms of (a) the problems in control of articulators and (b) the fact that it is an inherent feature independent to phonetic environment; (2) become familiar with the identification functions in discrimination of typical/atypical articulation of an acoustic cue; and (3) understand the difference between phonetic contrast and distortion contrast in speech perception. [Copyright &y& Elsevier]

Summary: An acoustic analysis of whispered consonants in comparison to normally phonated consonants was conducted in time and intensity domains. Consonant duration and average root mean square intensity were measured for six speakers in both articulation modes. Each of 25 Serbian consonants (C) was sited between the vowel /a/ forming a syllable of /aCa/ type. Such a syllable was placed in initial, medial, and final position in the carrier sentence. Results showed that whispered consonants have a prolonged duration of about 10% on average (statistically significant, ANOVA test), and that the unvoiced consonants have a smaller time dimension extension (5.8%) than voiced ones (15.3%). Examination at subphonemic level showed that there is no difference in voice-onset-time and affrication duration in unvoiced plosives and affricates, in both whispered and phonated mode of articulation, but the difference is significant for voiced ones. Analysis of consonant duration versus place of articulation showed that palatal place is most sensitive in the process of whispering. In all experiments, the results are very consistent with respect to the subjects and test material (Pearson''s correlation was between 0.6 and 0.9). In intensity domain, all unvoiced consonants in whispered mode of articulation have almost unchanged intensity in comparison to phonated mode (the difference is maximum 3.5dB). On the contrary, voiced consonants in the whispered mode were reduced in intensity by as much as 25dB, as nasals and semivowels. Average intensity of whispered consonants is lowered by 12dB in comparison to phonated ones, and does not depend on syllabic position inside the sentences. [Copyright &y& Elsevier]

The present study investigated the perceptual discrimination between typical (normal) acoustic features duration and intensity in word-initial fricative consonant /ʃ/ and atypical duration (shortened or lengthened) and intensity (weak or strong). Perceptual task distortion contrast was performed in each dimension, duration and intensity, as well as in two-dimensional interaction. Stimulus was the word /ʃuma/ with fricative /ʃ/ in the initial position. Duration and intensity of /ʃ/ were synthesized from typicaly pronounced stimuli in 16-step continuum from 135 ms to 300 ms and from −18 dB to +4.5 dB, respectively. Eight speech-language pathologies (SLPs) performed perceptual task distortion contrast in these two dimensions. Results are presented in the form of identification functions. First experiment tested two-choice decision and three-choice decision tasks in identification of distortion in time domain (shortened/normal/lengthened frictive), second experiment analyzed distortion in intensity domain (weak/normal/strong frictive), third experiment examined mutual interaction of distortion in duration and intensity, and forth experiment analyzed two-dimensional perception of simultaneous distortion in both domains, time and intensity. The results indicate regions of typical and atypical values for duration and intensity of fricative /ʃ/, in Serbian language, and also indicate that distortion in intensity domain does not influence perception of distortion in time domain, except in the case of weak friction. In two-dimensional perception of normal fricative /ʃ/, the dimension duration has higher perceptual significance over dimension intensity, whereas in perception of distorted fricative the dimension intensity has higher perceptual significance over dimension duration. In two-dimensional perception SLPs reported that they selectively recognize at first the dominant distortion and then another distortion. [ABSTRACT FROM AUTHOR]