Isolated Word Recognition Research Articles

Parallel Reservoir Computing Using Optical Amplifiers

Reservoir computing (RC), a computational paradigm inspired on neural systems, has become increasingly popular in recent years for solving a variety of complex recognition and classification problems. Thus far, most implementations have been software-based, limiting their speed and power efficiency. Integrated photonics offers the potential for a fast, power efficient and massively parallel hardware implementation. We have previously proposed a network of coupled semiconductor optical amplifiers as an interesting test case for such a hardware implementation. In this paper, we investigate the important design parameters and the consequences of process variations through simulations. We use an isolated word recognition task with babble noise to evaluate the performance of the photonic reservoirs with respect to traditional software reservoir implementations, which are based on leaky hyperbolic tangent functions. Our results show that the use of coherent light in a well-tuned reservoir architecture offers significant performance benefits. The most important design parameters are the delay and the phase shift in the system's physical connections. With optimized values for these parameters, coherent semiconductor optical amplifier (SOA) reservoirs can achieve better results than traditional simulated reservoirs. We also show that process variations hardly degrade the performance, but amplifier noise can be detrimental. This effect must therefore be taken into account when designing SOA-based RC implementations.

Noise Speech Recognition Based on Compressive Sensing

In this paper, we proposed a new noise speech recognition method based on the compressive sensing theory. Through compressive sensing, our method increases the anti-noise ability of speech recognition system greatly, which leads to the improvement of the recognition accuracy. According to the experiments, our proposed method achieved better recognition performance compared with the traditional isolated word recognition method based on DTW algorithm.

Acoustic Model Adaptation Based on Tensor Analysis of Training Models

We present a tensor analysis of acoustic models comprising various speakers in multiple noise conditions, and its application to the new speaker and environment adaptation for speech recognition. The bases used in adaptation are constructed by decomposing the training models in the state, feature dimension, speaker, and noise spaces using multilinear singular value decomposition. The isolated-word recognition experiment demonstrated the effectiveness of the proposed method, showing better performance than eigenvoice in the babble and factory floor noises for the adaptation data longer than approximately 20 s.

Morphological priming during reading: Evidence from eye movements

We report an eye movement experiment that investigated whether prior exposure to morphologically related and unrelated primes influenced processing of a target word that appeared later in the same sentence. Prime-target pairs had a semantically transparent (e.g., marshy-marsh) or only an apparent morphological relationship (e.g., secretary-secret), or were morphologically unrelated but overlapped in orthography (e.g., extract-extra). Reading times for target words revealed facilitation effects in measures of both early and late processing only for targets that followed semantically transparent morphological primes, providing evidence of semantically mediated priming between words read normally in a sentence. In addition, an increase in target word skipping and in regressions from a posttarget region when targets followed primes rather than control words, regardless of the morphological relationship between the words, suggests that prime-target orthographic overlap influences parafoveal processing of target words. We discuss our findings in relation to morphological priming during isolated word recognition and the process of lexical identification during reading.

F068 T and C levels in Nucleus CI24RE (CA) & CI24R(ST)/CI24M cochlear implant recipients: do they follow a common pattern?

Objectives: The purpose of this study was to find a correlation between cochlear implant performances in phoneme discrimination and activity of the brain-stem.Methods: Electrically-evoked auditory brain-stem responses (EABRs) and speech recognition performances were measured in 17 patients implanted with an MXM Digisonic DX10 cochlear implant. Speech recognition performances without lip-reading were tested using lists of isolated French words containing 3 phonemes.Results: The results indicated statistically significant correlations between phoneme correct-identification scores and the following EABR variables: wave V latency, wave II–V latency interval and wave III–V latency interval. These results, indicate that up to about 48% of the variance in isolated word recognition without lip-reading can be accounted for by EABR variables.Conclusion: The quality of brain-stem functioning influences central processes in phoneme discrimination.

Speech Recognition by Wavelet Analysis

ABSTRACT In an effort to provide a more efficient representation of the speech signal, the application of the wavelet analysis is considered. This research presents an effective and robust method for extracting features for speech processing. Based on the time‐frequency multi‐resolution property of wavelet transform, the input speech signal is decomposed into various frequency channels. The major issues concerning the design of this Wavelet based speech recognition system are choosing optimal wavelets for speech signals, decomposition level in the DWT, selecting the feature vectors from the wavelet coefficients. More specifically automatic classification of various speech signals using the DWT is described and compared using different wavelets. Finally, wavelet based feature extraction system and its performance on an isolated word recognition problem are investigated. For the classification of the words, three layered feed forward network is used. General Terms

Optimized discriminative transformations for speech features based on minimum classification error

Feature extraction is an important component of pattern classification and speech recognition. Extracted features should discriminate classes from each other while being robust to environmental conditions such as noise. For this purpose, several feature transformations are proposed which can be divided into two main categories: data-dependent transformation and classifier-dependent transformation. The drawback of data-dependent transformation is that its optimization criteria are different from the measure of classification error which can potentially degrade the classifier’s performance. In this paper, we propose a framework to optimize data-dependent feature transformations such as PCA (Principal Component Analysis), LDA (Linear Discriminant Analysis) and HLDA (Heteroscedastic LDA) using minimum classification error (MCE) as the main objective. The classifier itself is based on Hidden Markov Model (HMM). In our proposed HMM minimum classification error technique, the transformation matrices are modified to minimize the classification error for the mapped features, and the dimension of the feature vector is not changed. To evaluate the proposed methods, we conducted several experiments on the TIMIT phone recognition and the Aurora2 isolated word recognition tasks. The experimental results show that the proposed methods improve performance of PCA, LDA and HLDA transformation for mapping Mel-frequency cepstral coefficients (MFCC).

Adaptive Reliability Measure and Optimum Integration Weight for Decision Fusion Audio-visual Speech Recognition

Audio-visual speech recognition (AVSR) using acoustic and visual signals of speech has received attention recently because of its robustness in noisy environments. An important issue in decision fusion based AVSR system is the determination of appropriate integration weight for the speech modalities to integrate and ensure better performance under various SNR conditions. Generally, the integration weight is calculated from the relative reliability of two modalities. This paper investigates the effect of reliability measure on integration weight estimation and proposes a genetic algorithm (GA) based reliability measure which uses optimum number of best recognition hypotheses rather than N best recognition hypotheses to determine an appropriate integration weight. Further improvement in recognition accuracy is achieved by optimizing the above measured integration weight by genetic algorithm. The performance of the proposed integration weight estimation scheme is demonstrated for isolated word recognition (incorporating commonly used functions in mobile phones) via multi-speaker database experiment. The results show that the proposed schemes improve robust recognition accuracy over the conventional unimodal systems, and a couple of related existing bimodal systems, namely, the baseline reliability ratio-based system and N best recognition hypotheses reliability ratio-based system under various SNR conditions.

Dynamic Finger Movement Tracking and Voice Commands Based Smart Wheelchair

many disable persons are able to operate a joystick to control the electric wheelchair but on the other hand, many people do not have the legerdemain necessary to operate a joystick. Moreover, many of them face difficulty to avoid obstacles. The aim of this paper is to implement a multi-control system to control the movement of the wheelchair by integrating finger movement tracking system, a small vo c abulary speaker dependent based word recognition system and a group of monitoring sensors in order to avoid obstacles. Moreover, a joystick control system is also implemented in order to facilitate the patients that are able to use the joystick. The theme is accomplished by using a micro-controller with digital signal processor for isolated word recognition and sets of sensors. The results are also given at the end of the paper.

A new GA optimised Reliability Ratio based integration weight estimation scheme for decision fusion Audio-Visual Speech Recognition

Audio-Visual Speech Recognition (AVSR) using acoustic and visual signals of speech have received attention recently because of its robustness in noisy environments. An important issue in decision fusion AVSR systems is the determination of appropriate integration weight for better performance. A new Genetic Algorithm (GA) based scheme to obtain an appropriate integration weight is proposed here. The performance of the proposed scheme is demonstrated for commonly used mobile functions isolated word recognition via multi-speaker database experiment. The results show that the proposed scheme improves robust recognition accuracy over conventional unimodal systems and other related bimodal systems, namely, Reliability ratio and Neural Network based AVSR systems.

The Design of Isolated Word Recognition Based on ARM9

Nowadays,the theory and technology of speech recognition have made great progress,but in actual usage,they worked not very well.For some usage on some certain conditions ,we designed a small list, isolated word recognizer. With this system, we chosed ARM9 as the processor,MFCC as the characteristic parameters of speech sequences and adopted DTW to do pattern matching, As a result,it achieved hign rate of recognition and performed perfectly.

Efficient MLP constructive training algorithm using a neuron recruiting approach for isolated word recognition system

This paper describes an efficient constructive training algorithm using a Multi Layer Perceptron (MLP) neural network dedicated for Isolated Word Recognition (IWR) systems. Incremental training procedure was employed and this approach was based on novel hidden neurons recruiting for a single hidden-layer. During Neural Network (NN) training phase, the number of pronunciation samples extracted from the Training Data (TD) was sequentially increased. Optimal structure of the NN classifier with optimized TD size was obtained using this proposed MLP constructive training algorithm. Isolated word recognition system based on MLP neural network was then constructed and tested for recognizing ten words extracted from TIMIT database. Mel Frequency Cepstral Coefficient (MFCC) feature extraction method was employed including energy, first and second derivative coefficients. A proposed Frame-by-Frame Neural Network (FFNN) classification method was explored and compared with the Conventional Neural Network (CNN) classification approach. Principal Component Analysis (PCA) technique was also investigated in order to reduce both TD size as well as recognition system complexity. Experimental results showed superior performance of the proposed FFNN classifier compared to the CNN counter part which was illustrated by the significant improvement obtained in terms of recognition rate.

음성 변환을 사용한 감정 변화에 강인한 음성 인식

본 논문에서는 인간의 감정 변화에 강인한 음성 인식 시스템을 구현하기 위하여 음성 변환 방법 중의 한가지인 주파수 와핑 방법을 사용한 연구를 수행하였다. 이러한 목표를 위하여 다양한 감정이 포함된 음성 데이터베이스를 사용하여 감정의 변화에 따라 음성의 스펙트럼이 변화한다는 것과 이러한 변화는 음성 인식 시스템의 성능을 저하시키는 원인 중의 하나임을 관찰하였다. 본 논문에서는 이러한 음성의 변화를 감소시키는 방법으로 주파수 와핑을 학습 과정에 사용하는 방법을 제안하여 감정 변화에 강인한 음성 인식 시스템을 구현하였고 성도 길이 정규화 방법을 사용한 방법과 성능을 비교하였다. HMM을 사용한 단독음 인식 실험에서 제안된 학습 방법은 사용하면 감정이 포함된 데이터에 대한 인식 오차가 기존 방법보다 감소되었다. This paper studied some methods which use frequency warping method that is the one of the speech transformation method to develope the robust speech recognition system for the emotional variation. For this purpose, the effect of emotional variations on the speech signal were studied using speech database containing various emotions and it is observed that speech spectrum is affected by the emotional variation and this effect is one of the reasons that makes the performance of the speech recognition system worse. In this paper, new training method that uses frequency warping in training process is presented to reduce the effect of emotional variation and the speech recognition system based on vocal tract length normalization method is developed to be compared with proposed system. Experimental results from the isolated word recognition using HMM showed that new training method reduced the error rate of the conventional recognition system using speech signal containing various emotions.

Isolated word recognition of silent speech using magnetic implants and sensors

There are a number of situations where individuals wish to communicate verbally but are unable to use conventional means—so called ‘silent speech’. These include speakers in noisy and covert situations as well as patients who have lost their voice as a result of a laryngectomy or similar procedure. This paper focuses on those who are unable to speak following a laryngectomy and assesses the possibility of speech recognition based on a magnetic implant/sensors system. Permanent magnets are placed on the tongue and lips and the changes in magnetic field resulting from movement during speech are monitored using a set of magnetic sensors. The sensor signals are compared to sets of pre-recorded templates using the dynamic time warping (DTW) method, and the best match is identified. Experimental trials are reported for subjects with intact larynx, typically using 500–1000 utterances used for speaker dependant training and testing. It is shown that recognition rates of over 90% are achievable for vocabularies of at least 57 isolated words: sufficient to drive command-and-control applications.

Use of center of gravity with the common vector approach in isolated word recognition

In this paper, the subspace based classifier, common vector approach (CVA), with the center of gravity (COG) method is used for isolated word recognition. Since the CVA classifier is sensitive to shifts through the time axis, endpoint detection becomes extremely important for the recognition of isolated words. The COG method eliminates the need for endpoint detection. The effects of the COG method and a classical endpoint detection algorithm on the recognition rates of isolated words are investigated. The experimental results show that the COG method yields slightly higher recognition rates than the endpoint detection method in the TI-digit database when CVA is used.

주파수 와핑을 이용한 감정에 강인한 음성 인식 학습 방법

This paper studied the training methods less affected by the emotional variation for the development of the robust speech recognition system. For this purpose, the effect of emotional variation on the speech signal and the speech recognition system were studied using speech database containing various emotions. The performance of the speech recognition system trained by using the speech signal containing no emotion is deteriorated if the test speech signal contains the emotions because of the emotional difference between the test and training data. In this study, it is observed that vocal tract length of the speaker is affected by the emotional variation and this effect is one of the reasons that makes the performance of the speech recognition system worse. In this paper, a training method that cover the speech variations is proposed to develop the emotionally robust speech recognition system. Experimental results from the isolated word recognition using HMM showed that propose method reduced the error rate of the conventional recognition system by 28.4% when emotional test data was used.

Assessing Multimodal Spoken Word-in-Sentence Recognition in Children With Normal Hearing and Children With Cochlear Implants

To examine multimodal spoken word-in-sentence recognition in children. Two experiments were undertaken. In Experiment 1, the youngest age with which the multimodal sentence recognition materials could be used was evaluated. In Experiment 2, lexical difficulty and presentation modality effects were examined, along with test-retest reliability and validity in normal-hearing children and those with cochlear implants. Normal-hearing children as young as 3.25 years and those with cochlear implants just under 4 years who have used their device for at least 1 year were able to complete the multimodal sentence testing. Both groups identified lexically easy words in sentences more accurately than lexically hard words across modalities, although the largest effects occurred in the auditory-only modality. Both groups displayed audiovisual integration with the highest scores achieved in the audiovisual modality, followed sequentially by auditory-only and visual-only modalities. Recognition of words in sentences was correlated with recognition of words in isolation. Preliminary results suggest fair-to-good test-retest reliability. The results suggest that children's audiovisual word-in-sentence recognition can be assessed using the materials developed for this investigation. With further development, the materials hold promise for becoming a test of multimodal sentence recognition for children with hearing loss.

A novel approach to HMM-based speech recognition systems using particle swarm optimization

The main core of HMM-based speech recognition systems is Viterbi algorithm. Viterbi algorithm uses dynamic programming to find out the best alignment between the input speech and a given speech model. In this paper, dynamic programming is replaced by a search method which is based on particle swarm optimization algorithm. The major idea is focused on generating an initial population of segmentation vectors in the solution search space and improving the location of segments by an updating algorithm. Several methods are introduced and evaluated for the representation of particles and their corresponding movement structures. In addition, two segmentation strategies are explored. The first method is the standard segmentation which tries to maximize the likelihood function for each competing acoustic model separately. In the next method, a global segmentation tied between several models and the system tries to optimize the likelihood using a common tied segmentation. The results show that the effect of these factors is noticeable in finding the global optimum while maintaining the system accuracy. The idea was tested on an isolated word recognition and phone classification tasks and shows its significant performance in both accuracy and computational complexity aspects.

Cued Speech automatic recognition in normal-hearing and deaf subjects

This article discusses the automatic recognition of Cued Speech in French based on hidden Markov models (HMMs). Cued Speech is a visual mode which, by using hand shapes in different positions and in combination with lip patterns of speech, makes all the sounds of a spoken language clearly understandable to deaf people. The aim of Cued Speech is to overcome the problems of lipreading and thus enable deaf children and adults to understand spoken language completely. In the current study, the authors demonstrate that visible gestures are as discriminant as audible orofacial gestures. Phoneme recognition and isolated word recognition experiments have been conducted using data from a normal-hearing cuer. The results obtained were very promising, and the study has been extended by applying the proposed methods to a deaf cuer. The achieved results have not shown any significant differences compared to automatic Cued Speech recognition in a normal-hearing subject. In automatic recognition of Cued Speech, lip shape and gesture recognition are required. Moreover, the integration of the two modalities is of great importance. In this study, lip shape component is fused with hand component to realize Cued Speech recognition. Using concatenative feature fusion and multi-stream HMM decision fusion, vowel recognition, consonant recognition, and isolated word recognition experiments have been conducted. For vowel recognition, an 87.6% vowel accuracy was obtained showing a 61.3% relative improvement compared to the sole use of lip shape parameters. In the case of consonant recognition, a 78.9% accuracy was obtained showing a 56% relative improvement compared to the use of lip shape only. In addition to vowel and consonant recognition, a complete phoneme recognition experiment using concatenated feature vectors and Gaussian mixture model (GMM) discrimination was conducted, obtaining a 74.4% phoneme accuracy. Isolated word recognition experiments in both normal-hearing and deaf subjects were also conducted providing a word accuracy of 94.9% and 89%, respectively. The obtained results were compared with those obtained using audio signal, and comparable accuracies were observed.

Comparison of HMM and DTW for Isolated Word Recognition System of Punjabi Language

Comparison of HMM and DTW for Isolated Word Recognition System of Punjabi Language