Continuous Hindi speech recognition using Gaussian mixture HMM

Speech production knowledge has been used to enhance the phonetic representation and the performance of automatic speech recognition (ASR) systems successfully. Representations of speech production make simple explanations for many phenomena observed in speech. These phenomena can not be easily analyzed from either acoustic signal or phonetic transcription alone. One of the most important aspects of speech production knowledge is the use of articulatory knowledge, which describes the smooth and continuous movements in the vocal tract. In this paper, we present a new articulatory model to provide available information for rescoring the speech recognition lattice hypothesis. The articulatory model consists of a feature front-end, which computes a voicing feature based on a spectral harmonics correlation (SHC) function, and a back-end based on the combination of deep neural networks (DNNs) and hidden Markov models (HMMs). The voicing features are incorporated with standard Mel frequency cepstral coefficients (MFCCs) using heteroscedastic linear discriminant analysis (HLDA) to compensate the speech recognition accuracy rates. Moreover, the advantages of two different models are taken into account by the algorithm, which retains deep learning properties of DNNs, while modeling the articulatory context powerfully through HMMs. Mandarin speech recognition experiments show the proposed method achieves significant improvements in speech recognition performance over the system using MFCCs alone.

Automatic lip-reading (ALR) is a challenging task and a significant amount of research has been devoted to this topic in recent years. However, continuous Russian speech recognition still remains a not well-investigated area. In this paper, we present the results of Russian visual speech recognition (VSR) system using pixel-based and advanced geometry-based features. A HAVRUS video database, comprising of 4000 utterances of continuous Russian speech, collected from 20 speakers, is used in this study. Pixel-based features (principal component analysis-based or PCA) and geometry-based features (active appearance model-based or AAM) were used for the feature representation, and a Gaussian mixture hidden Markov models (HMM) were used for classification. Our evaluation experiments show a significant improvement (up to 9%) in recognition accuracy by using proposed geometry-based features when compared to pixel-based PCA features. The combined VSR is planned for future studies to augment the performance of audio-based automatic speech recognition systems in human–robot interfaces.

In this paper, we propose a novel boosted mixture learning (BML) framework for Gaussian mixture HMMs in speech recognition. BML is an incremental method to learn mixture models for classification problem. In each step of BML, one new mixture component is calculated according to functional gradient of an objective function to ensure that it is added along the direction to maximize the objective function the most. Several techniques have been proposed to extend BML from simple mixture models like Gaussian mixture model (GMM) to Gaussian mixture hidden Markov model (HMM), including Viterbi approximation to obtain state segmentation, weight decay to initialize sample weights to avoid overfitting, combining partial updating with global updating of parameters and using Bayesian information criterion (BIC) for parsimonious modeling. Experimental results on the WSJ0 task have shown that the proposed BML yields relative word and sentence error rate reduction of 10.9% and 12.9%, respectively, over the conventional training procedure.

This paper aims to design and implement English digits speech recognition system using Matlab (GUI). This work was based on the Hidden Markov Model (HMM), which provides a highly reliable way for recognizing speech. The system is able to recognize the speech waveform by translating the speech waveform into a set of feature vectors using Mel Frequency Cepstral Coefficients (MFCC) technique This paper focuses on all English digits from (Zero through Nine), which is based on isolated words structure. Two modules were developed, namely the isolated words speech recognition and the continuous speech recognition. Both modules were tested in both clean and noisy environments and showed a successful recognition rates. In clean environment and isolated words speech recognition module, the multi-speaker mode achieved 99.5% whereas the speaker-independent mode achieved 79.5%. In clean environment and continuous speech recognition module, the multi-speaker mode achieved 72.5% whereas the speaker-independent mode achieved 56.25%. However in noisy environment and isolated words speech recognition module, the multi-speaker mode achieved 88% whereas the speaker-independent mode achieved 67%. In noisy environment and continuous speech recognition module, the multi-speaker mode achieved 82.5% whereas the speaker-independent mode achieved 76.67%. These recognition rates are relatively successful if compared to similar systems.

Accurate and computationally efficient means of recognizing continuous speech has been a subject of research in recent years. This paper reports the development of a small vocabulary, speaker independent continuous Malayalam speech recognition system based on Hidden Markov Models (HMMs). Continuous density HMM, which is used in this work to model phonemes, represents the general case where the observation probability density functions (pdfs) are continuous. The observation pdf is approximated using a Gaussian mixture density. Mel-frequency Cepstral Coefficients (MFCC) method is used to extract acoustic features from the input signal. To represent temporal variations in the speech signal, the first and second order derivatives of MFCC are added to the set of static parameters. The training and decoding are performed by the Baum-Welch and Viterbi algorithms respectively. The corpus contains naturally and continuously spoken sentences with multiple pronunciations and speaker variations. On evaluation the proposed system has produced promising results with 94.67% word accuracy and 93.33% sentence correct.

In this dissertation, a novel approach of integrating template matching with statistical modeling is proposed to improve continuous speech recognition. Hidden Markov Modeling (HMMs) has been the dominant approach in statistical speech recognition since it provides a principled way of jointly modeling speech spectral variations and time dynamics. However, HMMs have the shortcoming of assuming the observations being independent within each state, which makes it ineffective in modeling the details of speech temporal evolutions that are important for characterizing nonstationary speech sounds. Template-based methods make comparisons between a test pattern and the templates derived from training data, and therefore they are able to capture speech dynamics and time correlation of speech frames better than HMM based methods. However, template matching requires large memory space and computational time since feature vectors of training data need to be stored in computer memory for access at the recognition stage, which is difficult in large vocabulary continuous speech recognition (LVCSR). Our proposed approach takes advantages of both statistical modeling and template matching, which overcomes the weakness of conventional template-based method and is feasible for LVCSR. We use multiple Gaussian Mixture Model (GMM) indices to represent each frame of speech templates, and define the template unit to be context-dependent phone segments (triphone context). We also use phonetic decision trees borrowed from those commonly used in HMMs to tie triphone templates and predict triphones unseen in training data. Two local distances, log likelihood ratio (LLR) and Kullback-Leibler (KL) divergence, are proposed for dynamic time warping (DTW) based template matching. In order to reduce computational complexity and storage space, we propose methods of minimum distance template selection (MDTS) and maximum log-likelihood template selection (MLTS), and investigate a template compression method on top of template selection to further improve recognition performance. The template based methods were used to rescore lattices generated by baseline HMMs on the tasks of TIMIT continuous phone recognition and teleheath LVCSR and experimental results demonstrated that the proposed approach of integrating template matching with statistical modeling significantly improved recognition performances over the HMM baselines. The template selection methods also provided significant recognition accuracy improvements over the HMM baseline while largely reducing the computation and storage complexities. When all templates or MDTS were used, using the LLR local distance obtained better recognition performance than the KL divergence local distance. For MLTS and template compression, KL divergence local distance provided better performance than the LLR local distance, and the template compression method made further improvements over KL based MLTS. Since the templates were constructed based on the GMM indices extracted from HMM baselines, we also validated the effectiveness of the proposed template methods based on enhanced HMM baselines. Experimental results showed that LLR based all template method was able to consistently improve TIMIT phone recognition accuracies based on four enhanced HMM baselines. Prosodic features such as duration, energy, and pitch can reflect longer span information of speech than conventional single frame vectors but they have commonly been ignored by HMMs. Template based methods provide possibilities to conveniently integrate prosodic features into speech recognition, which has not been well studied in the past. In this dissertation, we investigate combining template based methods with the speech prosodic features of duration, energy and pitch to further improve speech recognition accuracy. The scores of prosodic information were computed by a GMM based method and a non-parametric method, and the prosodic scores were combined with the acoustic scores in triphone template matching. Experimental results obtained on the telehealth task showed that prosodic information had positive effects on vowel sound recognition.

Combined speech enhancement and auditory modelling for robust distributed speech recognition

Inherent features of the Bangla (widely used as Bengali) language like long and short vowels and many instances of allophones make it difficult to build a continuous speech recognizer for the language. Stress and accent vary in spoken Bangla language from region to region. But in formal read Bangla speech, stress and accents are ignored. There are three approaches to continuous speech recognition (CSR) based on the sub-word unit viz. word, phoneme and syllable. Pronunciation of words and sentences are strictly governed by set of linguistic rules. Many attempts have been made to build continuous speech recognizers for Bangla for small and restricted tasks. However, medium and large vocabulary CSR for Bangla is relatively new and not explored. In this paper, the authors have attempted for extracting local features (LFs) from a Bangla input speech for tri-phone based automatic speech recognition (ASR) method. The method comprises two stages, where the first stage extracts LFs from input speech and the final stage generates word strings based on trip hone hidden Markov models (HMMs). The objective of this research is to build a medium vocabulary trip hone based continuous speech recognizer for Bangla language by extracting LFs over mel frequency cepstral coefficients (MFCCs). In this experimentation using Bangla speech corpus prepared by us, the recognizer provides higher word accuracy, word correct rate as well as sentence correct rate for trained and tested sentences with fewer mixture components in HMMs.

This study analyzes the effect of stress in human and automatic stressed speech processing tasks for speech collected from non-professional speakers. The database of 33 keywords is collected under five stress conditions, namely, neutral, angry, happy, sad and Lombard from fifteen speakers. The first study is to understand the ability to identify stress by human and automatic speech processing. The average performance of human stress classification is 59.44%. The average performance of automatic stress classifier using Vector Quantization (VQ) and Hidden Markov model (HMM) is 54.65% and 56.02%, respectively. The second study has been done to understand the effect of stress in human and automatic speech recognition. The average performance of human stressed speech recognition is 99.60%. The automatic stressed speech recognition performance using VQ and HMM is 82.42% and 76.79%, respectively. Even in the non-professional speakers scenario, human performance is better than automatic processing. Also, automatic processing seem to show considerable degradation in performance that warrant development of new methods to handle stress information.

The aim of this paper is to exhibit a comparative case study of the conventional speech recognition GMM-HMM (Gaussian mixture model - hidden Markov model) architecture and the recent model based on deep neural networks. During years the GMM approach has controlled the speech recognition tasks, however it has been surpassed with the resurgence of artificial neural networks. To exemplify these acoustic modeling frameworks, a case study has been conducted by using the Kaldi toolkit, employing a personalized speaker-independent mid-vocabulary voice corpus for recognition of digit strings and personal name lists in latin spanish on a connected-words phone dialing task. The speech recognition accuracy obtained in the results shows a better word error rate by using the DNN acoustic modeling. A 20.71% relative improvement is obtained with DNN-HMM models (3.33% WER) in respect to the lowest GMM-HMM rate (4.20% WER).

Speech recognition of continuous speech will be greatly influenced by the word segmentation process of the speech input. Proper segmentation will result in better speech recognition. This study proposed automatic speech recognition by applying local adaptive thresholding in the segmentation process. The segmentation method used is an enhanced blocking block area method whose input is a spectrogram image of the speech signal. While the locally adaptive thresholding method used is the Niblack method which is the best method compared to other methods, namely Sauvola, Bradley, Guanglei Xiong, and Bernsen when applied to the enhanced blocking block area method. For the speech recognition process, using Mel-frequency cepstral coefficients (MFCC) as a feature extraction method and Hidden Markov Model (HMM) as speech recognition. The experimental results show that by using 400 sentences consisting of 80 testing data and 320 as training data and using K-fold cross-validation, the highest accuracy is 60,8%. This result has no significant difference with the use of global thresholding in the segmentation process.

A speaker's accent is the most important factor affecting the performance of automatic speech recognition (ASR) systems. This is due to the fact that accents vary widely, even within the same country or community. The reason may be attributed to the fuzziness between the boundaries of phoneme classes, a result of differences in a speaker's vocal tract and accent. In this paper, a new method of accent classification is proposed that is based on fuzzy Gaussian mixture models (FGMMs). The proposed method first uses a fuzzy clustering to fuzzily partition the data. In this way, fuzzy memberships to the cluster centres are determined by minimizing the distance between the cluster centres and feature vectors. Afterwards, a GMM classifier is trained by using the fuzzy Gaussian parameters to classify the speaker's accent. The experimental results show that the proposed method outperforms the Gaussian Mixture models, Vector Quantization modeling method, Hidden Markov Model, and Radial Basis Neural Networks.

In this paper a novel algorithm for Gaussian Selection (GS) of mixtures used in a continuous speech recognition system is presented. The system is based on hidden Markov models (HMM), using Gaussian mixtures with full covariance matrices as output distributions. The purpose of Gaussian selection is to increase the speed of a speech recognition system, without degrading the recognition accuracy. The basic idea is to form hyper-mixtures by clustering close mixtures into a single group by means of Vector Quantization (VQ) and assigning it unique Gaussian parameters for estimation. In the decoding process only those hyper-mixtures which are above a designated threshold are selected, and only mixtures belonging to them are evaluated, improving computational efficiency. There is no problem with the clustering and evaluation if overlaps between the mixtures are small, and their variances are of the same range. However, in real case, there are numerous models which do not fit this profile. A Gaussian selection scheme proposed in this paper addresses this problem. For that purpose, beside the clustering algorithm, it also incorporates an algorithm for mixture grouping. The particular mixture is assigned to a group from the predefined set of groups, based on a value aggregated from eigenvalues of the covariance matrix of that mixture using Ordered Weighted Averaging operators (OWA). After the grouping of mixtures is carried out, Gaussian mixture clustering is performed on each group separately.

Speech recognition can be defined as the process of converting voice signals into the ranks of the word, by applying a specific algorithm that is implemented in a computer program. The research of speech recognition in Indonesia is relatively limited. This paper has studied methods of feature extraction which is the best among the Linear Predictive Coding (LPC) and Mel Frequency Cepstral Coefficients (MFCC) for speech recognition in Indonesian language. This is important because the method can produce a high accuracy for a particular language does not necessarily produce the same accuracy for other languages, considering every language has different characteristics. Thus this research hopefully can help further accelerate the use of automatic speech recognition for Indonesian language. There are two main processes in speech recognition, feature extraction and recognition. The method used for comparison feature extraction in this study is the LPC and MFCC, while the method of recognition using Hidden Markov Model (HMM). The test results showed that the LPC method is better than MFCC in Indonesian language speech recognition.

In this paper we develop a Turkish speech recognition (SR) system using deep neural networks and compare it with the previous state-of-the-art traditional Gaussian mixture model-hidden Markov model (GMM-HMM) method using the same Turkish speech dataset and the same large vocabulary Turkish corpus. Nowadays most SR systems deployed worldwide and particularly in Turkey use Hidden Markov Models to deal with the speech temporal variations. Gaussian mixture models are used to estimate the amount at which each state of each HMM fits a short frame of coefficients which is the representation of an acoustic input. A deep neural network consisting of feed-forward neural network is another way to estimate the fit; this neural network takes as input several frames of coefficients and gives as output posterior probabilities over HMM states. It has been shown that the use of deep neural networks can outperform the traditional GMM-HMM in other languages such as English and German. The fact that Turkish language is an agglutinative language and the lack of a huge amount of speech data complicate the design of a performant SR system. By making use of deep neural networks we will obviously improve the performance but still we will not achieve better result than English language due to the difference in the availability of speech data. We present various architectural and training techniques for the Turkish DNN-based models. The models are tested using a Turkish database collected from mobile devices. In the experiments, we observe that the Turkish DNN-HMM system have decreased the word error rate approximately 2.5% when compared to the GMM-HMM traditional system.