Recognition Of Speech Signals Research Articles

Features of the Recognition of Speech Signals in Conditions of Vocal Competition in Health and in Impairments to Auditory Speech Function

Features of the Recognition of Speech Signals in Conditions of Vocal Competition in Health and in Impairments to Auditory Speech Function

Effect of water-waves on recognition of speech signals transmitted over a wireless optical communication channel

In many sea-exploration scenarios, a wireless optical communication link between an onshore station and a floating marine structure or between a floating buoy and an underwater vehicle is required. In order to establish high-quality wireless optical communication, the effect of sea-waves must be taken into consideration. In this work, we investigate the influence of surface water waves having various amplitudes, frequencies and waveforms on the quality of an analog speech signal transmitted over a wireless optical communication link. Such investigation is also carried out on an encrypted speech signal. The quality of the speech signals is determined by utilizing a deep learning neural network that is trained to recognize ten speech commands and produces classification probabilities. In our experiments, either the detector or the light source is non-stationary as a result of the water-waves, thus distorting the received analog signal. We also show a method of deriving the frequency of the water waves by using a remote tracking unit based on micro-electro-mechanical systems (MEMS) mirrors.

Features of speech signal recognition in conditions of vocal competition with normal hearing and with hearing or speech disorders

We studied the features of speech signals perception in conditions of vocal competition based on gender (male/female voice) in subjects of different ages and status of auditory and speech functions. Psychophysical measurements were carried out while simulating a “speech cocktail” situation by simultaneously pronounced different words by a male and female. The average fundamental voice frequency for the male speaker was 108 ± 5.92 Hz, for the female speaker – 185 ± 12.03 Hz. Both speakers were standard native speakers of the Russian language. Digital recordings of speech signals were equalized in intensity and compounded so that the total test stimulus was a synchronized mixture of words spoken by a male (M) and female (F) voice. Test speech signals were presented through headphones or through a speaker located in front of the auditor at a distance of 50 cm. The indicators of reaction time and the number of correct recognitions of words of the target speaker (M or F) were compared in 4 groups of subjects of different ages and hearing or speech status: adult subjects with normal hearing and speech (n = 35) and with hearing impaired (n = 26); schoolchildren with normotypical development (n = 26) and with speech disorders (n = 25). The results indicated a falling in the ability to segrerate target speech streams under conditions of vocal competition in the subjects with hearing and speech dysfunctions included in the comparative study. Differences in the perception of male and female voices in subjects with hearing loss and with speech problems were obtained. This result may have a biological and social basis. The data have practical significance for the development of a system of auditory and speech training and modern hearing aid technologies.

A method for extracting the semantic features of speech signal recognition based on empirical wavelet transform

The subject of this study is methods for improving the efficiency of semantic coding of speech signals. The purpose of this study is to develop a method for improving the efficiency of semantic coding of speech signals. Coding efficiency refers to the reduction of the information transmission rate with a given probability of error-free recognition of semantic features of speech signals, which will significantly reduce the required source bandwidth, thereby increasing the communication channel bandwidth. To achieve this goal, it is necessary to solve the following scientific tasks: (1) to investigate a known method for improving the efficiency of semantic coding of speech signals based on mel-frequency cepstral coefficients; (2) to substantiate the effectiveness of using the adaptive empirical wavelet transform in the tasks of multiple-scale analysis and semantic coding of speech signals; (3) to develop a method of semantic coding of speech signals based on adaptive empirical wavelet transform with further application of Hilbert spectral analysis and optimal thresholding; and (4) to perform an objective quantitative assessment of the increase in the efficiency of the developed method of semantic coding of speech signals in contrast to the existing method. The following scientific results were obtained during the study: a method of semantic coding of speech signals based on empirical wavelet transform is developed for the first time, which differs from existing methods by constructing a set of adaptive bandpass Meyer wavelet filters with further application of Hilbert spectral analysis to find the instantaneous amplitudes and frequencies of the functions of internal empirical modes, which will allow the identification of semantic features of speech signals and increase the efficiency of their coding; for the first time, it is proposed to use the method of adaptive empirical wavelet transform in the tasks of multiple-scale analysis and semantic coding of speech signals, which will increase the efficiency of spectral analysis by decomposing the high-frequency speech oscillation into its low-frequency components, namely internal empirical modes; the method of semantic coding of speech signals based on mel-frequency cepstral coefficients was further developed, but using the basic principles of adaptive spectral analysis with the help of empirical wavelet transform, which increases the efficiency of this method. Conclusions: We developed a method for semantic coding of speech signals based on empirical wavelet transform, which reduces the encoding rate from 320 to 192 bps and the required bandwidth from 40 to 24 Hz with a probability of error-free recognition of approximately 0.96 (96%) and a signal-to-noise ratio of 48 dB, according to which its efficiency is increased by 1.6 times as compared to the existing method. We developed an algorithm for semantic coding of speech signals based on empirical wavelet transform and its software implementation in the MATLAB R2022b programing language.

Augmented Reality for Realtime Speech Recognition

Choosing the proper word out of samples with similar acoustic structures is a significant difficulty in speech recognition research, according to a case study of a basic domain-based speech recognition solution in speech recognition augmented reality products for hearing impaired persons. Our goal is to increase transcription accuracy in local vocabulary situations like meetings and lectures. To construct real-time ”live subtitles” utilising a unique voice recognition mechanism by combining our approach in an augmented reality environment. This programme allows hearing impaired and deaf persons to view real-time augmented reality subtitles while listening to a talk/speech on a certain topic. In the model of a tree-structured Hidden Markov Model (HMM) for speech recognition, there are different types of networks and graphs involved These networks and graphs play a crucial role in modeling and representing the relationships, transitions, and probabilities involved in speech recognition using tree-structured HMMs. They enable efficient modeling, adaptation, and recognition of speech signals in real- world noisy environments.

A Multi-Level Circulant Cross-Modal Transformer for Multimodal Speech Emotion Recognition

Speech emotion recognition, as an important component of human-computer interaction technology, has received increasing attention. Recent studies have treated emotion recognition of speech signals as a multimodal task, due to its inclusion of the semantic features of two different modalities, i.e., audio and text. However, existing methods often fail in effectively represent features and capture correlations. This paper presents a multi-level circulant cross-modal Transformer (MLCCT) for multimodal speech emotion recognition. The proposed model can be divided into three steps, feature extraction, interaction and fusion. Self-supervised embedding models are introduced for feature extraction, which give a more powerful representation of the original data than those using spectrograms or audio features such as Mel-frequency cepstral coefficients (MFCCs) and low-level descriptors (LLDs). In particular, MLCCT contains two types of feature interaction processes, where a bidirectional Long Short-term Memory (Bi-LSTM) with circulant interaction mechanism is proposed for low-level features, while a two-stream residual cross-modal Transformer block is applied when high-level features are involved. Finally, we choose self-attention blocks for fusion and a fully connected layer to make predictions. To evaluate the performance of our proposed model, comprehensive experiments are conducted on three widely used benchmark datasets including IEMOCAP, MELD and CMU-MOSEI. The competitive results verify the effectiveness of our approach.

Acquisition of English Corpus Machine Translation Based on Speech Recognition Technology

In the present information age, with the rapid development of computer software and hardware technology and the mature manufacturing system of English-speaking enterprises, it is no longer impossible to use statistics for machine translation. The level and quality of machine translation are expected to meet human expectations. This explains the meaning and realization value of the acquisition of English corpus machine translation, introduces the basic principles of speech recognition technology, and combines the characteristics of the English language on the basis of the original Chinese speech recognition system, adopts the technical means of speech recognition, and leads to in-depth research. In the new era of machine translation acquisition of English corpus, we use the combination of LabVIEW and MATLAB to complete the collection, editing, feature extraction, and speech recognition of speech signals and use VQ pattern matching technology to realize the English recognition of a large number of short vocabulary and individuals. In the application part, we applied the classic LabVIEW technology to the speech recognition technology, which actually realized the idea of “software instead of hardware” and achieved better translation results. Experiments show that the accuracy rate of English machine translation can be as high as 94% when using speech recognition technology. According to the results, it takes about 0.2 seconds to complete machine translation for a 30-second speech, which is basically okay to achieve the effect of real-time translation.

An Evaluation on Speech Recognition Technology based on Machine Learning

Speech is the basic way of interaction between the listener to the speaker by voice or expression. Humans can easily understand the speakers' message, but machines can't understand the speaker's word. Nowadays, most of our lives are occupied by machines; but we can't interact with machines. The human brain, like machine learning technology, is essential for speech recognition to interact with machines to humans. The language used for speech recognition must be a global language, so English has been used in this paper. The machine learning methodology is used in a lot of assignments through the feature learning capability. The data modelling capability results attained supplementary than the performance of normal learning methodology. So, in this work, the speech signal recognition is based on a machine-learning algorithm to merge the speech features and attributes. As a result of voice as a bio-metric implication, the speech signal is converted into a significant element of speech improvement. A new speech and emotion recognition technology is introduced. In this paper, discriminated speaking technology are spotlighted on the feature extraction, improvement, segmentation and progression of speech emotion recognition. Initially, the trained RNN layer-based feature extraction is done to get the speech signal's high-level features. From the generated high-level features are used for generating the new speech feature for the capsule network. Finally, the obtained speech features and attribute features are combined into the same RNN with Caps Net framework through the fully connected network. The experimental result shows the improved proposed speech recognition algorithms accuracy with another state-of-the-art method.

Research on Speech Recognition Method in Multi Layer Perceptual Network Environment

Aiming at the problem of low recognition rate in speech recognition methods, a speech recognition method in multi-layer perceptual network environment is proposed. In the multi-layer perceptual network environment, the speech signal is processed in the filter by using the transfer function of the filter. According to the framing process, the speech signal is windowed and framing processed to remove the silence segment of the speech signal. At the same time, the average energy of the speech signal is calculated and the zero crossing rate is calculated to extract the characteristics of the speech signal. By analyzing the principle of speech signal recognition, the process of speech recognition is designed, and the speech recognition in multi-layer perceptual network environment is realized. The experimental results show that the speech recognition method designed in this paper has good speech recognition performance

Gender Recognition of Speech Signal using KNN and SVM

Human speech is formulated by the vibration of vocal cord of human being consisting of different sounds for speaking, singing, expressing emotions and ideas, etc .The vocal cord being an important source of sound with human voice as a part of human sound synthesis [3] [5]. Differences in human speech for male and female are due to physiological difference such as vocal fold thickness or vocal tract length. The vocal tracts of male are mostly longer in comparison to vocal tracts of female. With advancing technology and more and more usage of human computer interaction (HCI) based systems, the speech processing plays an important role in enhancing the HCI system [2] [4]. The reason for development of gender recognition system is its used in gender based virtual assistants, telephonic surveys and voice controlled automatic systems.

Correction to: Emotion recognition of speech signal using Taylor series and deep belief network based classification

In the online published article, corresponding author has missed to update the following affiliation “Research scholar. Sathyabama Institute of Science and Technology, Chennai, India”. The current affiliation is given below.

WITHDRAWN: English vocabulary retrieval and recognition based on FPGA and machine learning

Abstract Digital audio spoken vocabulary retrieval and recognition have been developed in recent years. This growth has prompted extensive research into technology and reliable indexing. The English vocabulary content retrieval requires a combination of audio and speech processing technology and information retrieval. A study of English content search initially investigated units of planning to have a similar audio signal structure, and the subsequent focus outside the audio, naturally, more informal oral content. Shifted to generate conversational settings and the volume provided by this study outlines the component technical discipline of the relationship between lexical speech signal recognition and user interaction issues. The aim is for researchers with a background in speech technology to seek a deeper understanding of whether these fields are integrated into supporting research and development to solve the core challenges of vocabulary and content search. This study describes a content-based search method for Machine Learning based Neural Network classifiers that retrieve relevant documents efficiently and accurately. This overcomes the system limitations of characterization based arrangements regarding restricted vocabulary and preparing information accessibility. To analysis the system, it has designed a one-time learning neural network for classification and achieves better accuracy results.

Speech signal analysis of alzheimer's diseases in farsi using auditory model system.

In recent years, extensive studies have been conducted on the diagnosis of Alzheimer's disease (AD) using the non-invasive speech signal recognition method. In this study, Farsi speech signals were analyzed using the auditory model system (AMS) in order to recognize AD. For this purpose, after the pre-processing of the speech signals and utilizing AMS, 4D outputs as function of time, frequency, rate, and scale range were obtained. The AMS outcomes were averaged in term of time to analyze the rate-frequency-scale for both groups, Alzheimer's and healthy control subjects. Thereafter, the maximum of spectral and temporal modulation and frequency were extracted to classify by the support vector machine (SVM). The SVM achieves higher promising recognition accuracy with compare to prevalent approaches in the field of speech processing. The acceptable results demonstrate the applicability of the proposed algorithm in non-invasive and low-cost recognizing Alzheimer's only using the few extracted features of the speech signal.

Current Aspects of Speech Disfluency Research: Scientific Review of the International Workshop DiSS 19 (Budapest) Issues

Disfluency in spontaneous speech is currently a subject of study of specialists working in different fields of knowledge. Different external manifestations of disfluency (hesitation pauses, sound prolongations, pause fillers, articulatory perseverations and lexico-grammatical repetitions, self-corrections, breaks, nonverbal means of information transfer, etc.) are being investigated. They turn out to be a convenient tool for revealing and monitoring the peculiarities of cognitive processes with the help of explicit clearly registered signals occurring in speech under the influence of a number of extralinguistic factors such as the communicative situation, the type of speech (monologic or dialogic), the language of communication (L1 or L2), the emotional state of the speaker, the age, the social status, the diseases impairing speech and mental activity, and others. Further investigation of disfluency makes it possible to solve both a number of fundamental problems connected with modeling of cognitive coding and decoding speech processes and applied tasks connected with adoption of research findings in such fields as developmental pedagogy, psychology, medicine, foreign language training, translation, automatic recognition of speech signal, etc. Up to now, a sufficient number of empirical investigations have been carried out providing a basis for working out particular models which will make it possible, in the long run, to create the overall model of disfluency in spontaneous speech. Conferences and workshops undoubtedly play an important role in uniting the efforts of specialists in this sphere. One of them is the international workshop “Disfluency in Spontaneous Speech (DiSS)” that was first held in 1999. The current problems that were discussed by the participants of the workshop (production and perception speech models, age and clinical factors of disfluency, special difficulties in foreign speech production, including translation, speech technology) may be a useful reference point for researchers working on the issue.

EEG-based emotion recognition using an improved radial basis function neural network

The most advanced human–computer interaction is to make computers, like humans, capable of intelligent perception, judgment, and feedback. In the interaction, the emotions of the interactors can be identified to make intelligent measures. Emotion recognition mainly includes the recognition of speech, facial expressions, text, gestures, and physiological signals. Among them, emotional recognition in physiological signals is the most authentic. Since the EEG signal is a comprehensive reflection of the activities of many neurons in the brain in the cerebral cortex and can directly reflect brain activity, the EEG is rich in useful information. Therefore, this article uses EEG signals for the study of emotion recognition. First, the EEG is collected, preprocessed, and feature extracted; then an improved radial basis function neural network (I-RBF-NN) algorithm is used to process the EEG data; finally, the experimental results obtained by different classification models are compared and analyzed. The experimental results show that the I-RBF-NN proposed in this paper is better than other comparison algorithms for emotion recognition of EEG.

Emotion recognition of speech signal using Taylor series and deep belief network based classification

In the recent years, one of the multidisciplinary research areas attracting the researchers is emotion recognition. It is an important and challenging process to be achieved in emotional interaction. Accordingly, this work introduces the emotion recognition system by proposing the Taylor series based Deep Belief Network (Taylor-DBN). The noise present in the speech signal is removed through the speech enhancement process and then, subjected to the feature extraction. The features, such as tonal power ratio, Multiple Kernel Mel Frequency Cepstral Coefficients (MKMFCC) parameters, and the spectral flux are extracted and provided as the training input to the proposed Taylor-DBN classifier for identifying the emotions present in the signal. The experimentation is done with the help of the Berlin database, real database 1, and the real database 2. The experimental datasets contain the speech signals from different domain and language, and the performance of the proposed Taylor-DBN has shown minimal variations in each domain, and thus, the proposed model is suitable for various domains. The proposed Taylor-DBN outclassed other comparative models with Accuracy, False Acceptance Rate (FAR), and False Rejection Rate (FRR) values of 0.97, 0.0135, and 0.0165, respectively.

An artificial intelligence based speech model for linguistics teaching

The linguistic artificial intelligence teaching model can be assisted by the intelligent speech recognition model. The traditional speech recognition algorithm has certain problems, so it cannot effectively eliminate speech noise. Based on the advantages of the linguistics teaching model, this article combines the linguistics teaching model and the artificial intelligence model to build an artificial intelligence assisted teaching model that can be used for classroom teaching. Moreover, this study improves the traditional algorithm and constructs an artificial intelligence linguistics teaching model based on the improved algorithm. The filtering part of noise includes preliminary filtering of speech signals based on the short-term energy detection method, and further detection and recognition of preliminary filtering speech signals based on the artificial intelligence model detection method. After these two steps of filtering and recognition, the voice file is sent to the client for processing and control. In addition, this study set up a control experiment to analyze the performance of the model. The research results show that the algorithm in this paper has a certain effect.

Telugu Speech Recognition on TRI-SPECTRAL and DNN Techniques

This Research focus on the recognition of speech signals for Telugu language. The data of Telugu language considered is in isolated format. 10 isolated words are considered which are frequently spoken and recognized. Advanced technique named Tri spectral technique and DNN is used for this recognition. Tri spectral is a feature extraction technique. DNN is a feature classification technique. This research can be used in many interfacing systems which helps the humans to interact with the hardware or software systems easily. Design of ASR (“Automatic Speech Recognition System”) deals with many parameters which should finally conclude with promising recognition results. This techniques used in this research has given a better result with the accuracy of approximately 96.27%.

Itakura–Saito Divergence as an Element of the Information Theory of Speech Perception

On the basis of the information theory of speech perception, a substantiation of the symmetric form of the Itakura–Saito divergence as the minimum decision-making statistics of the asymptotically optimal algorithm for recognition of speech signals at the basic, phonetic level of their processing is given. The conclusions of the theoretical study are confirmed by the results of an experiment. It is shown that the use of the synthesized algorithm can significantly improve the accuracy and reliability of the automatic recognition of the most problematic phonetic units.

Recognition and Processing of Speech Signals Using Neural Networks

This paper provides an overview of recent approaches to deep learning as applied to speech processing tasks, primarily for automatic speech recognition, but also text-to-speech and speaker, language and emotion recognition. The focus is on efficient methods, addressing issues of accuracy, computation, storage, and delay. The discussion puts the speech processing tasks in the broader context of pattern recognition, comparing with signals other than speech. It also compares machine learning with other recent methods of speech analysis, e.g., hidden Markov models. The paper emphasizes a thorough understanding of the choices made in analyzing and interpreting speech signals. It minimizes use of mathematics and is aimed at non-experts; the references provide needed detail for those interested.