Features For Speech Recognition Research Articles

Appearance and shape-based hybrid visual feature extraction: toward audio–visual automatic speech recognition

Nowadays, audio–visual automatic speech recognition (AV-ASR) is an emerging field of research, but there is still lack of proper visual features for visual speech recognition. Visual features are mainly categorized into shape based and appearance based. Based on the different information embedded in shape and appearance features, this paper proposes a new set of hybrid visual features which lead to a better visual speech recognition system. Pseudo-Zernike Moment (PZM) is calculated for shape-based visual feature while Local Bnary Pattern-three orthogonal planes (LBP-TOP) and Discrete Cosine Transform (DCT) are calculated for the appearance-based feature. Moreover, our proposed method also gathers global and local visual information. Thus, the objective of the proposed system is to embed all this visual information into a compact features set. Here, for audio speech recognition, the proposed system uses Mel-frequency cepstral coefficients (MFCC). We also propose a hybrid classification method to carry out all the experiments of AV-ASR. Artificial Neural Network (ANN), multiclass Support Vector Machine (SVM) and Naive Bayes (NB) classifiers are used for classifier hybridization. It is shown that the proposed AV-ASR system with a hybrid classifier significantly improves the recognition rate.

An Experimental Analysis of Speech Features for Tone Speech Recognition

Recently Automatic Speech Recognition (ASR) has been successfully integrated in many commercial applications. These applications are performing significantly well in relatively controlled acoustical environments. However, the performance of an Automatic Speech Recognition system developed for non-tonal languages degrades considerably when tested for tonal languages. One of the main reason for this performance degradation is the non-consideration of tone related information in the feature set of the ASR systems developed for non-tonal languages. In this paper we have investigated the performance of commonly used feature for tonal speech recognition. A model has been proposed for extracting features for tonal speech recognition. A statistical analysis has been done to evaluate the performance of proposed feature set with reference to the Apatani language of Arunachal Pradesh of North-East India, which is a tonal language of Tibeto-Burman group of languages.

Aplikasi Pembelajaran Hiragana Bahasa Jepang Berbasis Android Menggunakan Speech Recognition

Alternatif mempelajari bahasa Jepang adalah dengan menggunakan mobile learning. Berdasarkan perbandingan yang dilakukan terhadap 7 aplikasi pembelajaran bahasa Jepang yang populer di playstore, belum terdapat aplikasi pembelajaran bahasa Jepang yang memiliki fitur speech recognition untuk pengucapan kalimat Jepang. Pada penelitian ini dirancang aplikasi pembelajaran hiragana Jepang untuk kalimat menggunakan speech recognition. Aplikasi ini menggunakan google cloud speech-to-text untuk fitur speech recognition. Algoritma yang diterapkan pada aplikasi ini adalah levenshtein distance untuk membandingkan inputan suara pengguna dengan teks pada database. Aplikasi berbasis android untuk pengguna dan web untuk admin. Berdasarkan hasil pengujian akurasi dan keberhasilan deteksi, jarak ideal aplikasi menerima inputan suara pengguna adalah 50 cm pada lingkungan tanpa noise dengan keberhasilan deteksi sebesar 100%. Pada pengujian sebelum dan sesudah menggunakan aplikasi terjadi peningkatan pengetahuan kosa kata sebesar 44.48% setelah pengguna menggunakan aplikasi. Berdasarkan pengujian usabilitas menggunakan SUS diperoleh hasil 73,01 yang menunjukkan bahwa aplikasi termasuk kedalam kategori acceptable.

Incorporating finer acoustic phonetic features in lexicon for Hindi language speech recognition

Lexicon, used in speech recognition, is the list of words and corresponding pronunciation using a standard acoustic notation. Currently ILSL12 phone-set is widely used for Indian Language speech recognition. However, it has limitations to represent features of the speech like, voicing, fricatives, etc. that are found in Indian Languages. This paper addresses this limitation by considering the voiced and unvoiced features for Hindi speech recognition by incorporating finer representations at the time of lexicon expansion. The approach is tested for Hindi word recognition and has shown significant improvement in WER.

Audio–Visual (Multimodal) Speech Recognition System Using Deep Neural Network

Abstract Recognizing the voice of a person by a smart device has become the need of the hour. Our cell phones are equipped with such features that converts speech to text, receives a command as a voice instruction instead of screen touch or pressing of a button. There are numerous intelligent machines such as automatic cars, robots, smart phones that use speech recognition features. These features are also very useful for people who are disabled and cannot use the smart devices with their hands. There are numerous techniques available to implement a speech recognition system. The need of such a system is to accurately recognize human voice at any circumstances, because the accent of each human differs for a single language, and also, the human voice is never devoid of noise and other emotional elements. So, the aim of such systems is to recognize human voices of different accents, even in a noisy environment. Because of this, an effective and robust speech recognition system has become important for machine and human interaction. Much voice-recognizing software depends purely on audio to recognize a voice, and one cannot guarantee that the audio will be noise free. In order to improve the accuracy and performance of traditional speech recognition system in a noisy environment, a new, nontraditional approach called a multimodal speech recognition system that uses both the audio and video features such as lip movement facial expression to recognize a speech is used. This article discusses the advantage of using video features along with the audio features with an experiment that proves it.

A method for feature extraction based on SVD and machine learning

By studying the shortcomings of feature, which extracted from Radar-Cross Section(RCS),using mathematical and statistical method, using the idea of extracting abstract features in image recognition and speech recognition by artificial intelligence for reference[2][3]. This paper explores the possibility of extracting abstract features of target’s RCS sequence, and proposes an abstract feature extraction method of RCS sequence based on singular value decomposition(SVD) feature decomposition. Because of the poor interpretability of abstract features, four different machine learning algorithms are used to classify the extracted features. The experimental results show that the machine learning algorithm can classify different types of spatial objects with high accuracy, which shows that the RCS features of different spatial objects can be characterized by abstract features.

Effect of emotionally toned Malay language sounds on the brain: a NIRS analysis

Speech recognition features such as emotion have always been involved in human communication. With the recent developments in the communication methods, researchers have investigated artificial and emotional intelligence to improve communication. This has led to the emergence of affective computing, which deals with processing information pertaining to human emotions. This study aims to determine positive influence of language sounds containing emotion on brain function for improved communication. Twenty-seven college-age Japanese subjects with no prior exposure to the Malay language listened to emotionally toned and emotionally neutral sounds in the Malay language. Their brain activities were measured using near-infrared spectroscopy (NIRS) as they listened to the sounds. A comparison between different NIRS signals revealed that emotionally toned language sounds had a greater impact on brain areas associated with attention and emotion. On the contrary, emotionally neutral Malay sounds affected brain areas involved in working memory and language processing. These results suggest that emotionally-charged sounds initiate listeners’ attention and emotion recognition even when the listeners do not understand the language. The ability to interpret emotions presents challenges in computer systems and robotics; therefore, we hope that our results can be used for the development of computational models of emotion for autonomous robot research in the field of communication.

Training of reduced-rank linear transformations for multi-layer polynomial acoustic features for speech recognition

The use of higher-order polynomial acoustic features can improve the performance of automatic speech recognition (ASR). However, dimensionality of polynomial representation can be prohibitively large, making acoustic model training using polynomial features infeasible for large vocabulary ASR systems. This paper presents a multi-layer polynomial training framework for acoustic modeling, which recursively expands the acoustic features into their second-order polynomial feature space. After each expansion the dimensionality of resultant features is reduced by a linear transformation. Experimental results obtained for two large-vocabulary continuous speech recognition tasks show that the proposed method outperforms conventional mixture models. More recently the acoustic modelling community has shifted its focus to deep neural networks. We also train multi-layer polynomial features in a similar way: allowing backpropagation and using mean-normalized stochastic gradient descent algorithm. This has led to encouraging results. Specifically, appending a sigmoid-based feed-forward deep neural network with a final polynomial layer has resulted in significant word error rate improvement.

A Study of Speech Feature Extraction Based on Manifold Learning

Manifold learning is a nonlinear data dimension reduction method. It can look for the essence of things from the observed phenomena, and find the inherent law of data. Traditional MFCC feature will lead a slower learning speed on account of it has high dimension and useless noise. Therefore, a speech feature extraction method based on manifold learning is proposed. Firstly, we use the manifold learning dimension reduction algorithm for the dimension reduction of Mel features and then for vowels classification. In order to further demonstrate the effectiveness of manifold learning feature in speech recognition, we propose a fusion speech feature extraction method and apply it to the identification of Chinese isolated words. Experiments prove that the fusion feature extraction method has achieved a better result than that of traditional MFCC feature extraction method.

Enhanced shalat and wudhu learning media through speech recognition application

Nowadays, learning methods on the procedure of shalat and wudhu are found everywhere, both formal institutions, non formal and through learning media such as books and smartphones. However, these media are considered not interest children to learn the procedures of shalat and wudhu. This work proposes an enhanced shalat and wudhu learning media through speech recognition feature on input method of searching search material of shalat and wudhu. This technology allows a device to recognize and understand the word that spoke by someone to digitizing words and matching those digital signals to a specific pattern that was stored in the media stored. Knuth Morris Pratt was chosen as an algorithm used for the matching of keywords with the material for shalat and wudhu. In limited test, this app was using by 10 kids, with 7-9 years old. The result, 7 of 10 kids interested to learn shalat and wudhu using this app. The searching modul with Knutt Morris Pratt has working properly with 80% accuration.

Robust Speech Recognition Using Adaptive Noise Cancellation

This paper introduces the adaptive noise cancellation technique for the noise reduction in Robust Automatic Speech Recognition. The adaptive noise cancellation is used as front-end stage to enhance the extracted features for speech recognition under noisy conditions. More specifically, the Constrained Stability Least Mean Square (CS-LMS) algorithm which is a member of the family of adaptive filters has been applied. The Hidden Markov Model based Tool Kit (HTK) is used for training and testing the Automatic Speech Recognizer system. The result obtained shows that the application of adoptive filtering at the front-end enhances the performance of the system in noisy conditions while the CS-LMS algorithm gives the most superior performance among the family of LMS algorithms.

Pitch-Normalized Acoustic Features for Robust Children's Speech Recognition

In this letter, the effectiveness of recently reported SMAC (Spectral Moment time–frequency distribution Augmented by low-order Cepstral) features has been evaluated for robust automatic speech recognition (ASR). The SMAC features consist of normalized first central spectral moments appended with low-order cepstral coefficients. These features have been designed for achieving robustness to both additive noise and the pitch variations. We have explored the SMAC features in severe pitch mismatch ASR task, i.e., decoding of children's speech on adults’ speech trained ASR system. In those tasks, the SMAC features are still observed to be sensitive to pitch variations. Toward addressing the same, a simple spectral smoothening approach employing adaptive-cepstral truncation is explored prior to the computation of spectral moments. With the proposed modification, the SMAC features are noted to achieve enhanced pitch robustness without affecting their noise immunity. Furthermore, the effectiveness of the proposed features is explored in three dominant acoustic modeling paradigms and varying data conditions. In all the cases, the proposed features are observed to significantly outperform the existing ones.

Seismic facies analysis based on speech recognition feature parameters

Seismic facies analysis plays an important role in seismic stratigraphy. Seismic attributes have been widely applied to seismic facies analysis. One of the most important steps is to optimize the most sensitive attributes with regard to reservoir characteristics. Using different attribute combinations in multidimensional analyses will yield different solutions. Acoustic waves and seismic waves propagating in an elastic medium follow the same law of physics. The generation process of a speech signal based on the acoustic model is similar to the seismic data of the convolution model. We have developed the mel-frequency cepstrum coefficients (MFCCs), which have been successfully applied in speech recognition, as feature parameters for seismic facies analysis. Information about the wavelet and reflection coefficients is well-separated in these cepstrum-domain parameters. Specifically, information about the wavelet mainly appears in the low-domain part, and information about the reflection coefficients mainly appeared in the high-domain part. In the forward model, the seismic MFCCs are used as feature vectors for synthetic data with a noise level of zero and 5%. The Bayesian network is used to classify the traces. Then, classification accuracy rates versus different orders of the MFCCs are obtained. The forwarding results indicate that high accuracy rates are achieved when the order exceeds 10. For the real field data, the seismic data are decomposed into a set of MFCC parameters. The different information is unfolded in the parameter maps, enabling the interpreter to capture the geologic features of the target interval. The geologic features presented in the three instantaneous attributes and coherence can also be found in the MFCC parameter maps. The classification results are in accordance with the paleogeomorphy of the target interval as well as the known wells. The results from the synthetic data and real field data demonstrate the information description abilities of the seismic MFCC parameters. Therefore, using the speech feature parameters to extract information may be helpful for processing and interpreting seismic data.

Combining Binaural and Cortical Features for Robust Speech Recognition

The segregation of concurrent speakers and other sound sources is an important ability of the human auditory system, but is missing in most current systems for automatic speech recognition ASR, resulting in a large gap between human and machine performance. This study combines processing related to peripheral and cortical stages of the auditory pathway: A physiologically motivated binaural model estimates the positions of moving speakers to enhance the desired speech signal. Second, signals are converted to spectro-temporal Gabor features that resemble cortical speech representations and which have been shown to improve ASR in noisy conditions. Spectro-temporal Gabor features improve recognition results in all acoustic conditions under consideration compared with Mel-frequency cepstral coefficients. Binaural processing results in lower word error rates WERs in acoustic scenes with a concurrent speaker, whereas monaural processing should be preferred in the presence of a stationary masking noise. In-depth analysis of binaural processing identifies crucial processing steps such as localization of sound sources and estimation of the beamformer's noise coherence matrix, and shows how much each processing step affects the recognition performance in acoustic conditions with different complexity.

Wavelet based feature combination for recognition of emotions

Abstract In this paper, authors tried to develop reduced combinational features for emotional speech recognition. The spectral/cepstral features like wavelet coefficient, LPCC (linear prediction cepstral coefficient) and MFCC (mel-frequency cepstral coefficient) are used as baseline features. The frequency variation with respect to time has been evaluated properly using wavelet coefficients. Again, MFCC and LPCC features are derived from wavelets, so that the exact information of the emotion has been fetched and are used as features. The features are reduced with vector quantization method and used in radial basis function network (RBFNN) classifier. The feature sets are also combined and tested in the classifier. This piece of work deals with five emotions as angry, fear, happy, disgust and neutral. These are tested for Berlin (EMO-DB) and Surrey Audio-Visual Expressed Emotion (SAVEE) database. The proposed frequency based decomposition and combination choice of features show excellent result and it is exhibited in result section.

Phoneme Set Design Based on Integrated Acoustic and Linguistic Features for Second Language Speech Recognition

Phoneme Set Design Based on Integrated Acoustic and Linguistic Features for Second Language Speech Recognition

An Extraction Method of Acoustic Features for Speech Recognition

An Extraction Method of Acoustic Features for Speech Recognition

A new variance-based approach for discriminative feature extraction in machine hearing classification using spectrogram features

Machine hearing is an emerging research field that is analogous to machine vision in that it aims to equip computers with the ability to hear and recognise a variety of sounds. It is a key enabler of natural human–computer speech interfacing, as well as in areas such as automated security surveillance, environmental monitoring, smart homes/buildings/cities. Recent advances in machine learning allow current systems to accurately recognise a diverse range of sounds under controlled conditions. However doing so in real-world noisy conditions remains a challenging task. Several front–end feature extraction methods have been used for machine hearing, employing speech recognition features like MFCC and PLP, as well as image-like features such as AIM and SIF. The best choice of feature is found to be dependent upon the noise environment and machine learning techniques used. Machine learning methods such as deep neural networks have been shown capable of inferring discriminative classification rules from less structured front–end features in related domains. In the machine hearing field, spectrogram image features have recently shown good performance for noise-corrupted classification using deep neural networks. However there are many methods of extracting features from spectrograms. This paper explores a novel data-driven feature extraction method that uses variance-based criteria to define spectral pooling of features from spectrograms. The proposed method, based on maximising the pooled spectral variance of foreground and background sound models, is shown to achieve very good performance for robust classification.

Speaker-dependent features for spontaneous speech recognition

Speaker-dependent features for spontaneous speech recognition

Generative Model-Driven Feature Learning for dysarthric speech recognition

Recognition of speech uttered by severe dysarthric speakers needs a robust learning technique. One of the commonly used generative model-based classifiers for speech recognition is a hidden Markov model. Generative model-based classifiers do not do well for overlapping classes and due to insufficient training data. Dysarthric speech is normally partial or incomplete that leads to improper learning of temporal dynamics. To overcome these issues, we focus on learning features for dysarthric speech recognition that involves recognizing the sequential patterns of varying length utterances. We propose a Generative Model-Driven Feature Learning based discriminative framework that maps the sequence of feature vectors to fixed dimension vector spaces induced by the generative models. The discriminative classifier is built in that vector space. The proposed HMM-based fixed dimensional vector representation provides better discrimination for dysarthric speech than the conventional HMM. We examine the performance of the proposed method to recognize the isolated utterances from the UA-Speech database. The recognition accuracy of the proposed model is better than the conventional hidden Markov model-based approach.