Abstract Automatic Speech Recognition Research Articles

Challenges for Designing of Children Speech Corpora: A State-of-the-Art Review

Automatic Speech Recognition (ASR) is the use of computer hardware and software-based techniques to identify and process human voices. These systems used data from both male and female speakers. The majority of commercial ASR systems available on adult speech are working efficiently. Speech data collected from both male and female speakers were used in these systems. In recent decades, ASR systems for children have emerged, such as reading tutors, aids for foreign language learning, and computer games. Child ASR systems are essential but poorly understood in the field of computer speech recognition. The child data collection is a very complex task. Child corpus is not available publicly, and variability of children speakers and ASR developed for a particular age group is not suitable for other age groups. These are some of the reasons for less and ineffective child ASR systems. However, the non-availability of child corpus publicly is a primary reason for ineffective child ASR. Designing and developing child corpus is a very tedious task. Therefore, the primary focus of this state-of-the-art review is to discuss various challenges encountered while designing and developing child corpus.

Automatic speech recognition in the booth

Abstract Automatic Speech Recognition (ASR) has been proposed as a means to enhance state-of-the-art computer-assisted interpreting (CAI) tools and to allow machine-learning techniques to enter the workflow of professional interpreters. In this article, we test the usefulness of real-time transcription with number highlighting of a source speech for simultaneous interpreting using InterpretBank ASR. The system’s precision is high (96%) and its latency low enough to fit interpreters’ ear–voice span (EVS). We evaluate the potential benefits among first-time users of this technology by applying an error matrix and by investigating the users’ subjective perceptions through a questionnaire. The results show that the ASR provision improves overall performance for almost all number types. Interaction with the ASR support is varied and participants consult it for just over half of the stimuli. The study also provides some evidence of the psychological benefits of ASR availability and of overreliance on ASR support.

Lithuanian Speech Recognition Using Purely Phonetic Deep Learning

Automatic speech recognition (ASR) has been one of the biggest and hardest challenges in the field. A large majority of research in this area focuses on widely spoken languages such as English. The problems of automatic Lithuanian speech recognition have attracted little attention so far. Due to complicated language structure and scarcity of data, models proposed for other languages such as English cannot be directly adopted for Lithuanian. In this paper we propose an ASR system for the Lithuanian language, which is based on deep learning methods and can identify spoken words purely from their phoneme sequences. Two encoder-decoder models are used to solve the ASR task: a traditional encoder-decoder model and a model with attention mechanism. The performance of these models is evaluated in isolated speech recognition task (with an accuracy of 0.993) and long phrase recognition task (with an accuracy of 0.992).

Automatic speech recognition: A primer for speech-language pathology researchers

Automatic speech recognition (ASR) is increasingly becoming an integral component of our daily lives. This trend is in large part due to recent advances in machine learning, and specifically in deep learning, that have led to accurate ASR across numerous tasks. This has led to renewed interest in providing technological support to populations whose speech patterns are atypical, including identifying the presence of a specific pathology and its severity, comparing speech characteristics before and after a surgery and enhancing the quality of life of individuals with speech pathologies. The purpose of this primer is to bring readers with relatively little technical background up to speed on fundamentals and recent advances in ASR. It presents a detailed account of the anatomy of modern ASR, with examples of how it has been used in speech-language pathology research.

Chhattisgarhi speech corpus for research and development in automatic speech recognition

Automatic speech recognition (ASR) is a computerized interface which allows humans to communicate with machine in a way of its natural conversation. ASR has wide range of applications in various fields such as language development in young children, telecommunications, as an assistive device for hearing impaired etc. Performance of ASR system is greatly influenced by the database used for its implementation. In this paper, we are discussing about building a speech corpus for a rare but important Indian dialect Chhattisgarhi. This speech corpus consists of 100 unique isolated words and four speech scripts aggregating 67 sentences, recorded from total 478 native speakers. These words were selected from English to Chhattisgarhi dictionary published by Chhattisgarh Rajbhasha Aayog and scripts from Chhattisgarhi literature and newspaper articles. This dataset has been collected travelling over 60% geographical area of the Chhattisgarh state. Finally, a valuable speech corpus for the first time have been prepared for Chhattisgarhi with an aim to enhance the speech research. The successful extermination of speech recognition for both isolated and continuous speech samples have been demonstrated on the prepared database.

Implementation of Context-Dependent Acoustic Modelling for Sanskrit Vocabulary

Automatic speech recognition (ASR) is used for accurate and efficient conversion of speech signal into a text message. The area of ASR is being discussed since past few decades and significant advancement is being observed periodically on the ASR and language spoken systems. However, there are many technological hurdles yet to reach flexible solutions that satisfies user. This is because of many factors such as environmental noise, paucity of robustness to speech variations (foreign accents, sociolinguistics, gender and speaking rate), spontaneous or freestyle speech and others. To realise the ubiquitous adoption of speech technology, there is a need to bridge the space between what speech recognition technologies can convey and what human need from it. To make it up, technology must deliver robust and high-recognition accuracy near to man-like performance so it demands to focus on the challenges in speech technology. Generally, speech signal is taken as input, and it is processed at front end to extract features and then computed at back end using the Gaussian mixture model (GMM). GMMmixtureselectionis quiteimportantdependingupon thesizeofdataset. Asforconcisevocabulary, use of triphone-based acoustic modelling exhibits good result, the same has been implemented for Sanskrit language.

Toward completely automated vowel extraction: Introducing DARLA

AbstractAutomatic Speech Recognition (ASR) is reaching further and further into everyday life with Apple’s Siri, Google voice search, automated telephone information systems, dictation devices, closed captioning, and other applications. Along with such advances in speech technology, sociolinguists have been considering new methods for alignment and vowel formant extraction, including techniques like the Penn Aligner (Yuan and Liberman 2008) and the FAVE automated vowel extraction program (Evanini et al. 2009; Rosenfelder et al. 2011). With humans transcribing audio recordings into sentences, these semi-automated methods can produce effective vowel formant measurements (Labov et al. 2013). But as the quality of ASR improves, sociolinguistics may be on the brink of another transformative technology: large-scale, completely automated vowel extraction without any need for human transcription. It would then be possible to quickly extract vowels from virtually limitless hours of recordings, such as YouTube, publicly available audio/video archives, and large-scale personal interviews or streaming video. How far away is this transformative moment? In this article, we introduce a fully automated program called DARLA (short for “Dartmouth Linguistic Automation,”http://darla.dartmouth.edu), which automatically generates transcriptions with ASR and extracts vowels using FAVE. Users simply upload an audio recording of speech, and DARLA produces vowel plots, a table of vowel formants, and probabilities of the phonetic environments for each token. In this paper, we describe DARLA and explore its sociolinguistic applications. We test the system on a dataset of the US Southern Shift and compare the results with semi-automated methods.

Optimum integration weight for decision fusion audio-visual speech recognition

Automatic speech recognition (ASR) technologies have been successfully applied to several real world applications. But, still there exist several problems that need to be solved for wider application of the technologies. One problem is noise-robustness of recognition performance. Recently, audio-visual speech recognition (AVSR) has received attention as a solution to this problem. In this, visual speech information is used together with acoustic signal for speech recognition in noisy environments. This paper presents a new decision fusion AVSR system, in which the classifier's decision is optimised using genetic algorithm (GA) optimisation technique. Hence, the optimally fused decision fusion AVSR system produces robust recognition accuracy at all SNR conditions. For evaluating the performance of the proposed scheme, the recognition results are compared with those of an equal weight bimodal AVSR system and with another state-of-the-art method, namely, compression and Mel sub-band spectral subtraction (CMSBS)-based noise compensation method for speech recognition in noise. Further, to show the effectiveness of the proposed optimisation method, the recognition results are compared with those of a similar method called directed grid search method, which also optimises the integration weight against the recognition accuracy.

Statistical analysis of features and classification of alphasyllabary sounds in Kannada language

Automatic speech recognition (ASR) for a given audio file is a challenging task due to the variations in the type of speech input. Variations may be the environment, language spoken, emotions of the speaker, age/gender of speaker etc. The two main steps in ASR are converting the audio file into features and classifying it appropriately. Basic unit of speech sound is phoneme and the list of such phoneme is language dependent. In Indian languages, basic unit of language is known as Akshara i.e the alphabet. It is known to be an alphasyllabary unit. In our work, we have analyzed the behavior of the acoustic features like, Mel frequency cepstral coefficients and linear predictive coding for various aksharas using techniques like, visualization, probability density function (pdf), Q---Q plot and F-ratio. The classifiers, support vector machine (SVM) and hidden Markov model (HMM) are used for classifying the recorded audio into corresponding aksharas. We have also compared the classification performance of HMM and SVM.

Hindi phoneme-viseme recognition from continuous speech

Automatic Speech Recognition (ASR) system performs well under restricted conditions but the performance degrades under noisy environment. Audio–visual features play an important role in ASR systems in the presence of noise. In this paper, Hindi phoneme recognition system is designed using audio-visual features. The Discrete Cosine Transform (DCT) features of the lip region integrated with Mel Frequency Cepstral Coefficient (MFCC) audio features are used to get better recognition performance under noisy environments. Colour intensity, hybrid method and Pseudo-Hue methods have been used for lip-localisation approach with Linear Discriminant Analyser (LDA) as a classifier. Recognition performance using Pseudo-Hue method proved best among all the methods.

Network and Embedded Applications of Automatic Speech Recognition

ASR (Automatic Speech Recognition) is one of key technologies in the upcoming Ubiquitous Computing and Ambient Intelligence. In this paper, first, the surveys on processing devices such as microprocessors and memories, and on communication infrastructure, especially wireless communication infrastructure relating to ASR are reported. Second, the embedded version of CSR (Continuous Speech Recognition) software for the mobile environmental use of ASR is reported. As the devices, RISC based microprocessors, semiconductor memories, and HDD are summarized. For the communication infrastructure, mobile communications and wireless LANs are described. Finally, implementation results of the free CSR software called Julius on the T-engineTM consisting of an SH-4A mi-croprocessor are reported.

CLASSIFICATION OF MALAYSIAN VOWELS USING FORMANT BASED FEATURES

Automatic speech recognition (ASR) has made great strides with the development of digital signal processing hardware and software, especially using English as the language of choice. Despite of all these advances, machines cannot match the performance of their human counterparts in terms of accuracy and speed, especially in case of speaker independent speech recognition. In this paper, a new feature based on formant is presented and evaluated on Malaysian spoken vowels. These features were classified and used to identify vowels recorded from 80 Malaysian speakers. A back propagation neural network (BPNN) model was developed to classify the vowels. Six formant features were evaluated, which were the first three formant frequencies and the distances between each of them. Results, showed that overall vowel classification rate of these three formant combinations are comparatively the same but differs in terms of individual vowel classification.

Acoustic quality normalization for robust automatic speech recognition

Automatic speech recognition (ASR) system suffers from the variation of acoustic quality in an input speech. Speech may be produced in noisy environments and different speakers have their own way of speaking style. Variations can be observed even in the same utterance and the same speaker in different moods. All these uncertainties and variations should be normalized to have a robust ASR system. In this paper, we apply and evaluate different approaches of acoustic quality normalization in an utterance for robust ASR. Several HMM (hidden Markov model)-based systems using utterance-level, word-level, and monophone-level normalization are evaluated with HMM-SM (subspace method)-based system using monophone-level normalization for normalizing variations and uncertainties in an utterance. SM can represent variations of fine structures in sub-words as a set of eigenvectors, and so has better performance at monophone-level than HMM. Experimental results show that word accuracy is significantly improved by the HMM-SM-based system with monophone-level normalization compared to that by the typical HMM-based system with utterance-level normalization in both clean and noisy conditions. Experimental results also suggest that monophone-level normalization using SM has better performance than that using HMM.

Accelerating the generation of work measurement standards through automatic speech recognition: A laboratory study

Automatic Speech Recognition (ASR) is an emerging technology. As often is the case with new technologies, there is little guidance available on the considerations accompanying its incorporation into industrial settings. This paper presents the steps and factors to consider for successful integration of ASR systems into industrial environments. In particular, this work shows the successful application of ASR within the context of the generation of work measurement time standards. The generation of time standards is a time consuming process that has always slowed the work measurement task and increased its costs. It is shown that by using ASR the work analyst can develop more efficient, more reliable, detailed time estimates in real time, without the need for a keyboard or paper. In laboratory experiments, time reductions of 70% were demonstrated in generating time standards for common manual labour tasks. In two field studies, positive user satisfaction was expressed in the application of ASR for generating time standards.

Talking to Machines Today and Tomorrow: Designing for the User

Automatic speech recognition (ASR) is finding many new applications in both products and services. Designing good user interfaces with ASR poses a variety of challenges. This paper: — Addresses basic principles in designing for the user, — Describes the need for appropriate dialogues between the user and ASR-based systems, — Discusses how these user-machine transactions are changing due to improved ASR technology, and — Discusses the direction ASR should take to provide the kind of “natural” user-machine dialogue that is expected in Vision 2001.