Search

In this paper, the effect of prosody-modification-based data augmentation is explored in the context of automatic speech recognition (ASR). The primary motive is to develop ASR systems that are less affected by speaker-dependent acoustic variations. Two factors contributing towards inter-speaker variability that are focused on in this paper are pitch and speaking-rate variations. In order to simulate such an ASR task, we have trained an ASR system on adults’ speech and tested it using speech data from adult as well as child speakers. Compared to adults’ speech test case, the recognition rates are noted to be extremely degraded when the test speech is from child speakers. The observed degradation is basically due to large differences in pitch and speaking-rate between adults’ and children’s speech. To overcome this problem, pitch and speaking-rate of the training speech are modified to create new versions of the data. The original and the modified versions are then pooled together in order to capture greater acoustic variability. The ASR system trained on augmented data is noted to be more robust towards speaker-dependent variations. Relative improvements of 11.5% and 27.0% over the baseline are obtained on decoding adults’ and children’s speech test sets, respectively.

The presence of velopharyngeal dysfunction in individuals with cleft palate CP nasalizes the voiced stops. Due to this, voiced stops /b/, /d/, /g/ tend to be perceive like nasal consonants /m/, /n/, /ng/. In this work, a novel algorithm is proposed for the detection of nasalized voiced stops in CP speech using epoch-synchronous features. Speech regions corresponding to consonant and consonant-vowel transitions are segmented using the knowledge of glottal activity, syllable nucleus, low-frequency spectral dominance, and vowel onset point. The segmented regions are epoch-synchronously processed to analyze the spectral, spectro-temporal, excitation source, and periodicity characteristics of normal and nasalized voiced stops. Spectral and spectro temporal features are computed using single pole filter based time-frequency representation. The amplitude of Hilbert envelope of linear prediction residual, measured around the epoch is used to analyze the effect of nasalization on excitation source. Comparison of speech frames of successive inter-epoch intervals is carried out to analyze the periodicity characteristics. The proposed features are used to develop a support vector machine classifier for the classification of normal and nasalized voiced stops. Segmentation accuracy for the proposed knowledge based method is found to be better than the hidden Markov model based force-alignment approach. The detection rate of nasalized voiced stops is found to be high for the proposed epoch synchronous features than the conventional Mel-frequency cepstral coefficients.

In this paper, we present our efforts towards developing a robust automatic speaker verification (ASV) system for children when the domain-specific data is limited. For that purpose, we have studied the effect of in-domain and out-of-domain data augmentation. Several different combinations of data augmentation are studied in this work. Speed and pitch perturbation of children’s speech are employed for synthetically creating in-domain data to be used for augmentation. For out-of-domain data augmentation, on the other hand, adults’ speech is pooled together with children’s speech. At the same time, voice conversion (VC) is also applied on adults’ speech to alter the acoustic attributes. VC of adults’ speech makes it perceptually similar to that of children’s speech. The converted adults’ data is then used for augmentation. The ASV systems developed in this study employ x-vectors derived using a time-delay deep neural network. In addition to that, probabilistic linear discriminant analysis is used for scoring the performance. The explored methods of data augmentation are noted to reduce the equal error rate as well as minimum decision cost function by a large margin.

The objective of this paper is to demonstrate the significance of combining different features present in the glottal activity region for statistical parametric speech synthesis (SPSS). Different features present in the glottal activity regions are broadly categorized as F0, system, and source features, which represent the quality of speech. F0 feature is computed from zero frequency filter and system feature is computed from 2-D based Riesz transform. Source features include aperiodicity and phase component. Aperiodicity component representing the amount of aperiodic component present in a frame is computed from Riesz transform, whereas, phase component is computed by modeling integrated linear prediction residual. The combined features resulted in better quality compared to STRAIGHT based SPSS both in terms of objective and subjective evaluation. Further, the proposed method is extended to two Indian languages, namely, Assamese and Manipuri, which shows similar improvement in quality.

In the context of automatic speech recognition (ASR) systems, the front-end acoustic features should not be affected by signal periodicity (pitch period). Motivated by this fact, we have studied the role of pitch-synchronous spectrum estimation approach, referred to as TANDEM STRAIGHT, in this paper. TANDEM STRAIGHT results in a smoother spectrum devoid of pitch harmonics to a large extent. Consequently, the acoustic features derived using the smoothed spectra outperform the conventional Mel-frequency cepstral coefficients (MFCC). The experimental evaluations reported in this paper are performed on speech data from a wide range of speakers belonging to different age groups including children. The proposed features are found to be effective for all groups of speakers. To further improve the recognition of children's speech, the effect of vocal-tract length normalization (VTLN) is studied. The inclusion of VTLN further improves the recognition performance. We have also performed a detailed study on the effect of speaking-rate normalization (SRN) in the context of children's speech recognition. An SRN technique based on the anchoring of glottal closure instants estimated using zero-frequency filtering is explored in this regard. SRN is observed to be highly effective for child speakers belonging to different age groups. Finally, all the studied techniques are combined for effective mismatch reduction. In the case of children's speech test set, the use of proposed features results in a relative improvement of 21.6% over the MFCC features even after combining VTLN and SRN.

The objective of this paper is to present a detailed review of modelling various acoustic features employed in statistical parametric speech synthesis (SPSS). As reported in the literature, many ac...

Our efforts towards developing an automatic speaker verification (ASV) system for child speakers are presented in this paper. For the majority of the languages, children's speech data for training the ASV system is either unavailable (zero-resource) or very limited (low-resource). Under low- and zero-resource conditions, developing an ASV system becomes a very challenging problem. To overcome this issue, we have studied the effectiveness of in-domain and out-of-domain data augmentation in this work. Speed and pitch modifications of children's speech are employed for synthetically creating data in the case of in-domain data augmentation. On the other hand, a limited amount of adults' speech is used when out-of-domain data augmentation is performed. Using adults' speech leads to severe acoustic mismatch due to dissimilarity in the attributes of speech data from adult and child speakers. To address this drawback, speech data from adult speakers are subjected to voice conversion (VC) to alter the acoustic attributes. A cycle-consistent generative adversarial network is used in this work for voice conversion. Voice conversion renders adults' speech perceptually similar to children's speech. The voice converted adults' data can then be used for augmentation, ensuring that the acoustic mismatch is minimal. To study the effectiveness of proposed data augmentation techniques experimentally, x-vector-based ASV system architecture is employed. At the same time, the role of i-vector is also studied in this paper. As a consequence of data augmentation, both equal error rate and minimum decision cost function are reduced significantly in low- and zero-resource conditions. At the same time, employing i-vectors for modeling speaker characteristics is noted to be superior. Finally, we have also presented a detailed study on the effect of data augmentation with child speakers' age variation.

A method to improve voicing decision using glottal activity features proposed for statistical parametric speech synthesis. In existing methods, voicing decision relies mostly on fundamental frequency F 0 , which may result in errors when the prediction is inaccurate. Even though F 0 is a glottal activity feature, other features that characterize this activity may help in improving the voicing decision. The glottal activity features used in this work are the strength of excitation (SoE), normalized autocorrelation peak strength (NAPS), and higher-order statistics (HOS). These features obtained from approximated source signals like zero-frequency filtered signal and integrated linear prediction residual. To improve voicing decision and to avoid heuristic threshold for classification, glottal activity features are trained using different statistical learning methods such as the k-nearest neighbor, support vector machine (SVM), and deep belief network. The voicing decision works best with SVM classifier, and its effectiveness is tested using the statistical parametric speech synthesis. The glottal activity features SoE, NAPS, and HOS modeled along with F 0 and Mel-cepstral coefficients in Hidden Markov model and deep neural network to get the voicing decision. The objective and subjective evaluations demonstrate that the proposed method improves the naturalness of synthetic speech.

In this paper, we study the role of speaking-rate adaptation (SRA) of automatic speech recognition (ASR) systems. The performance of an ASR system is reported to degrade when the speaking-rate is either too fast or too slow. In order to simulate such a situation, an ASR system was trained on adults' speech and used for transcribing speech data from adult as well as child speakers. Earlier studies have shown that, speaking-rate is significantly lower in the case of children when compared to adults. Consequently, the recognition performance for children's speech was noted to be very poor in contrast to adults' speech. To improve the recognition performance with respect to children's speech, speaking-rate was explicitly changed using time-scale modification (TSM). A recently proposed TSM approach based on fuzzy classification of spectral bins has been explored in this regard. The fuzzy-classification-based TSM technique is reported to be superior to state-of-the-art approaches. Effectiveness of the said TSM technique has not been studied yet in the context of ASR. The experimental studies presented in this paper show that SRA based on fuzzy classification results in a relative improvement of 30% over the baseline.

In the context of automatic speech recognition (ASR), the power spectrum is generally warped to the Mel-scale during front-end speech parameterization. This is motivated by the fact that, human perception of sound is nonlinear. The Mel-filterbank provide better resolution for low-frequency contents while a greater degree of averaging happens in the high-frequency range. The work presented in this paper aims at studying the role of linear, Mel and inverse-Mel filterbanks in the context of speech recognition. It is well known that, when speech data is from high-pitched speakers like children, there is a significant amount of relevant information in the high-frequency region. Hence, down-sampling the information in that range through Mel-filterbank reduces the recognition performance. On the other hand, employing inverse-Mel or linear-filterbanks are expected to be more effective in such cases. The same has been experimentally validated in this work. To do so, an ASR system is developed on adults' speech and tested using data from adult as well as child speakers. Significantly improved recognition rates are noted for children's as well adult females' speech when linear or inverse-Mel filterbank is used. The use of linear filters results in a relative improvement of 21% over the baseline.

Developmental dysphasia or specific language impairment (SLI) is a disorder that is known to delay the process of acquiring language skills in children without other disabilities. Approximately 5–7% of children in kindergarten group are affected with SLI as reported in literature. Boys are more prone to be affected by this disorder compared to girls. In this paper, we present our preliminary attempts towards detecting SLI in children using their speech data. In this regard, we have used Mel-frequency cepstral coefficients (MFCC) for front-end speech parameterization. We have also presented an analysis to show how MFCC features help in discriminating the healthy children from those affected with SLI. The MFCC feature vectors are then used to develop two-class classifiers for discriminating healthy children from those suffering from SLI. The said two-class classifiers are developed using extreme learning machine (ELM) trained and tested on speech data collected from healthy children as well as those affected with SLI. ELM are fast to train and are known to be quite effective even when the training data is sparse. For extracting utterance-level features to be given as input to the ELM, Gaussian posteriograms learned on frame-level acoustic features are used. Several different types of ELMs are explored in this work and the kernel ELM is noted to outperform the rest with an accuracy of 99.41%.

Accurate detection of pause boundary plays a major role in the segmentation of the speech corpus and improving the quality of speech synthesis. For pause modelling, we need to have pause tags in the training sentences. Manual tagging of pause is accurate but have the possibilities of missing out due to human error, and it is time-consuming. In this work, an automatic approach for marking the pause in the training corpus is proposed. During the training phase, after every word explicit pause (PAU) tags are added to represent a pause. Then, models for all phones including PAU are trained and re-alignment is performed. During re-alignment, each PAU boundary is corrected using three speech specific features namely, modulation spectrum energy, spectral peaks energy, and strength of excitation. The proposed approach gives a better result as compared to manual pause marking with less time complexity. It also improves the overall segmentation accuracy. The tagged label files are used for developing text-to-speech synthesis system using Hidden Markov Model based speech synthesis framework. Subjective evaluation is performed for various approaches used in tagging the pause. The experimental evaluation shows that accurate pause marking plays an important factor for improving the quality of synthesized speech in terms of naturalness and intelligibility.