Search results for: statistical models of speech recognition

Authors: Christian Arcos, Marley Vellasco, Abraham Alcaim

Abstract:

In this paper, we present a wavelet coefficients masking based on Local Binary Patterns (WLBP) approach to enhance the temporal spectra of the wavelet coefficients for speech enhancement. This technique exploits the wavelet denoising scheme, which splits the degraded speech into pyramidal subband components and extracts frequency information without losing temporal information. Speech enhancement in each high-frequency subband is performed by binary labels through the local binary pattern masking that encodes the ratio between the original value of each coefficient and the values of the neighbour coefficients. This approach enhances the high-frequency spectra of the wavelet transform instead of eliminating them through a threshold. A comparative analysis is carried out with conventional speech enhancement algorithms, demonstrating that the proposed technique achieves significant improvements in terms of PESQ, an international recommendation of objective measure for estimating subjective speech quality. Informal listening tests also show that the proposed method in an acoustic context improves the quality of speech, avoiding the annoying musical noise present in other speech enhancement techniques. Experimental results obtained with a DNN based speech recognizer in noisy environments corroborate the superiority of the proposed scheme in the robust speech recognition scenario.

Keywords: binary labels, local binary patterns, mask, wavelet coefficients, speech enhancement, speech recognition

Procedia PDF Downloads 197

Authors: Sheena Christabel Pravin, M. Palanivelan

Abstract:

Speech disfluencies are common in spontaneous speech. The primary purpose of this study was to distinguish linguistic disfluencies from stuttering disfluencies in bilingual Tamil–English (TE) speaking children. The secondary purpose was to determine whether their disfluencies are mediated by native language dominance and/or on an early onset of developmental stuttering at childhood. A detailed study was carried out to identify the prosodic and acoustic features that uniquely represent the disfluent regions of speech. This paper focuses on statistical modeling of repetitions, prolongations, pauses and interjections in the speech corpus encompassing bilingual spontaneous utterances from school going children – English and Tamil. Two classifiers including Hidden Markov Models (HMM) and the Multilayer Perceptron (MLP), which is a class of feed-forward artificial neural network, were compared in the classification of disfluencies. The results of the classifiers document the patterns of disfluency in spontaneous speech samples of school-aged children to distinguish between Children Who Stutter (CWS) and Children with Language Impairment CLI). The ability of the models in classifying the disfluencies was measured in terms of F-measure, Recall, and Precision.

Keywords: bi-lingual, children who stutter, children with language impairment, hidden markov models, multi-layer perceptron, linguistic disfluencies, stuttering disfluencies

Procedia PDF Downloads 190

Authors: Tusar Kanti Dash, Ganapati Panda

Abstract:

The evaluation of speech quality and intelligence is critical to the overall effectiveness of the Speech Enhancement Algorithms. Several intrusive and non-intrusive measures are employed to calculate these parameters. Non-Intrusive Evaluation is most challenging as, very often, the reference clean speech data is not available. In this paper, a novel non-intrusive speech evaluation measure is proposed using audio features derived from the Stockwell transform. These features are used with the Light Gradient Boosting Machine for the effective prediction of speech quality and intelligibility. The proposed model is analyzed using noisy and reverberant speech from four databases, and the results are compared with the standard Intrusive Evaluation Measures. It is observed from the comparative analysis that the proposed model is performing better than the standard Non-Intrusive models.

Keywords: non-Intrusive speech evaluation, S-transform, light GBM, speech quality, and intelligibility

Procedia PDF Downloads 228

Authors: M. Draman, S. Z. Muhamad Yassin, M. S. Alias, Z. Lambak, M. I. Zulkifli, S. N. Padhi, K. N. Baharim, F. Maskuriy, A. I. A. Rahim

Abstract:

The performance of Malay automatic speech recognition (ASR) system for the call centre environment is presented. The system utilizes Kaldi toolkit as the platform to the entire library and algorithm used in performing the ASR task. The acoustic model implemented in this system uses a deep neural network (DNN) method to model the acoustic signal and the standard (n-gram) model for language modelling. With 80 hours of training data from the call centre recordings, the ASR system can achieve 72% of accuracy that corresponds to 28% of word error rate (WER). The testing was done using 20 hours of audio data. Despite the implementation of DNN, the system shows a low accuracy owing to the varieties of noises, accent and dialect that typically occurs in Malaysian call centre environment. This significant variation of speakers is reflected by the large standard deviation of the average word error rate (WERav) (i.e., ~ 10%). It is observed that the lowest WER (13.8%) was obtained from recording sample with a standard Malay dialect (central Malaysia) of native speaker as compared to 49% of the sample with the highest WER that contains conversation of the speaker that uses non-standard Malay dialect.

Keywords: conversational speech recognition, deep neural network, Malay language, speech recognition

Procedia PDF Downloads 296

Authors: R. Ajgou, S. Sbaa, S. Ghendir, A. Chemsa, A. Taleb-Ahmed

Abstract:

The speech enhancement algorithm is to improve speech quality. In this paper, we review some speech enhancement methods and we evaluated their performance based on Perceptual Evaluation of Speech Quality scores (PESQ, ITU-T P.862). All method was evaluated in presence of different kind of noise using TIMIT database and NOIZEUS noisy speech corpus.. The noise was taken from the AURORA database and includes suburban train noise, babble, car, exhibition hall, restaurant, street, airport and train station noise. Simulation results showed improved performance of speech enhancement for Tracking of non-stationary noise approach in comparison with various methods in terms of PESQ measure. Moreover, we have evaluated the effects of the speech enhancement technique on Speaker Identification system based on autoregressive (AR) model and Mel-frequency Cepstral coefficients (MFCC).

Keywords: speech enhancement, pesq, speaker recognition, MFCC

Procedia PDF Downloads 389

Authors: Nabil Shawkat, Jamil Saquer

Abstract:

Social media has accelerated our ability to engage with others and eliminated many communication barriers. On the other hand, the widespread use of social media resulted in an increase in online hate speech. This has drastic impacts on vulnerable individuals and societies. Therefore, it is critical to detect hate speech to prevent innocent users and vulnerable communities from becoming victims of hate speech. We investigate the performance of different deep learning and machine learning algorithms on three different datasets. Our results show that the BERT model gives the best performance among all the models by achieving an F1-score of 90.6% on one of the datasets and F1-scores of 89.7% and 88.2% on the other two datasets.

Keywords: hate speech, machine learning, deep learning, abusive words, social media, text classification

Procedia PDF Downloads 102

Authors: Rita Chattopadhyay, Vivek Anand Thoutam

Abstract:

Speech is the most acceptable means of communication where we can quickly exchange our feelings and thoughts. Quite often, people can communicate orally but cannot interact or work with computers or devices. It’s easy and quick to give speech commands than typing commands to computers. In the same way, it’s easy listening to audio played from a device than extract output from computers or devices. Especially with Robotics being an emerging market with applications in warehouses, the hospitality industry, consumer electronics, assistive technology, etc., speech-based human-machine interaction is emerging as a lucrative feature for robot manufacturers. Considering this factor, the objective of this paper is to design the “Audio-Visual Co-Data Processing Pipeline.” This pipeline is an integrated version of Automatic speech recognition, a Natural language model for text understanding, object detection, and text-to-speech modules. There are many Deep Learning models for each type of the modules mentioned above, but OpenVINO Model Zoo models are used because the OpenVINO toolkit covers both computer vision and non-computer vision workloads across Intel hardware and maximizes performance, and accelerates application development. A speech command is given as input that has information about target objects to be detected and start and end times to extract the required interval from the video. Speech is converted to text using the Automatic speech recognition QuartzNet model. The summary is extracted from text using a natural language model Generative Pre-Trained Transformer-3 (GPT-3). Based on the summary, essential frames from the video are extracted, and the You Only Look Once (YOLO) object detection model detects You Only Look Once (YOLO) objects on these extracted frames. Frame numbers that have target objects (specified objects in the speech command) are saved as text. Finally, this text (frame numbers) is converted to speech using text to speech model and will be played from the device. This project is developed for 80 You Only Look Once (YOLO) labels, and the user can extract frames based on only one or two target labels. This pipeline can be extended for more than two target labels easily by making appropriate changes in the object detection module. This project is developed for four different speech command formats by including sample examples in the prompt used by Generative Pre-Trained Transformer-3 (GPT-3) model. Based on user preference, one can come up with a new speech command format by including some examples of the respective format in the prompt used by the Generative Pre-Trained Transformer-3 (GPT-3) model. This pipeline can be used in many projects like human-machine interface, human-robot interaction, and surveillance through speech commands. All object detection projects can be upgraded using this pipeline so that one can give speech commands and output is played from the device.

Keywords: OpenVINO, automatic speech recognition, natural language processing, object detection, text to speech

Procedia PDF Downloads 49

Authors: Saurabh Farkya, Govinda Surampudi

Abstract:

Optical Character Recognition is one of the current major research areas. This paper is focussed on recognition of Devanagari script and its sound generation. This Paper consists of two parts. First, Optical Character Recognition of Devnagari handwritten Script. Second, speech synthesis of the recognized text. This paper shows an implementation of support vector machines for the purpose of Devnagari Script recognition. The Support Vector Machines was trained with Multi Domain features; Transform Domain and Spatial Domain or Structural Domain feature. Transform Domain includes the wavelet feature of the character. Structural Domain consists of Distance Profile feature and Gradient feature. The Segmentation of the text document has been done in 3 levels-Line Segmentation, Word Segmentation, and Character Segmentation. The pre-processing of the characters has been done with the help of various Morphological operations-Otsu's Algorithm, Erosion, Dilation, Filtration and Thinning techniques. The Algorithm was tested on the self-prepared database, a collection of various handwriting. Further, Unicode was used to convert recognized Devnagari text into understandable computer document. The document so obtained is an array of codes which was used to generate digitized text and to synthesize Hindi speech. Phonemes from the self-prepared database were used to generate the speech of the scanned document using concatenation technique.

Keywords: Character Recognition (OCR), Text to Speech (TTS), Support Vector Machines (SVM), Library of Support Vector Machines (LIBSVM)

Procedia PDF Downloads 469

Authors: Khenfer-Koummich Fatima, Mesbahi Larbi, Hendel Fatiha

Abstract:

This paper presents a recognition system for isolated words like robot commands. It’s carried out by Time Delay Neural Networks; TDNN. To teleoperate a robot for specific tasks as turn, close, etc… In industrial environment and taking into account the noise coming from the machine. The choice of TDNN is based on its generalization in terms of accuracy, in more it acts as a filter that allows the passage of certain desirable frequency characteristics of speech; the goal is to determine the parameters of this filter for making an adaptable system to the variability of speech signal and to noise especially, for this the back propagation technique was used in learning phase. The approach was applied on commands pronounced in two languages separately: The French and Arabic. The results for two test bases of 300 spoken words for each one are 87%, 97.6% in neutral environment and 77.67%, 92.67% when the white Gaussian noisy was added with a SNR of 35 dB.

Keywords: TDNN, neural networks, noise, speech recognition

Procedia PDF Downloads 253

Authors: Thein Thein, Kalyar Myo San

Abstract:

Lip reading system is one of the different supportive technologies for hearing impaired, or elderly people or non-native speakers. For normal hearing persons in noisy environments or in conditions where the audio signal is not available, lip reading techniques can be used to increase their understanding of spoken language. Hearing impaired persons have used lip reading techniques as important tools to find out what was said by other people without hearing voice. Thus, visual speech information is important and become active research area. Using visual information from lip movements can improve the accuracy and robustness of a speech recognition system and the need for lip reading system is ever increasing for every language. However, the recognition of lip movement is a difficult task because of the region of interest (ROI) is nonlinear and noisy. Therefore, this paper proposes method to detect the accurate lips shape and to localize lip movement towards automatic lip tracking by using the combination of Otsu global thresholding technique and Moore Neighborhood Tracing Algorithm. Proposed method shows how accurate lip localization and tracking which is useful for speech recognition. In this work of study and experiments will be carried out the automatic lip localizing the lip shape for Myanmar consonants using the only visual information from lip movements which is useful for visual speech of Myanmar languages.

Keywords: lip reading, lip localization, lip tracking, Moore neighborhood tracing algorithm

Procedia PDF Downloads 328

Authors: Ahmad Alwosheel, Ahmed Alqaraawi

Abstract:

This work proposes an unsupervised text-independent probabilistic approach to recognize Quran reciter voice. It is an accurate approach that works on real time applications. This approach does not require a prior information about reciter models. It has two phases, where in the training phase the reciters' acoustical features are modeled using Gaussian Mixture Models, while in the testing phase, unlabeled reciter's acoustical features are examined among GMM models. Using this approach, a high accuracy results are achieved with efficient computation time process.

Keywords: Quran, speaker recognition, reciter recognition, Gaussian Mixture Model

Procedia PDF Downloads 352

Authors: Muhammad Imran Ahmad, Ruzelita Ngadiran, Mohd Nazrin Md Isa, Nor Ashidi Mat Isa, Mohd ZaizuIlyas, Raja Abdullah Raja Ahmad, Said Amirul Anwar Ab Hamid, Muzammil Jusoh

Abstract:

This paper presents two technique, local feature extraction using image spectrum and low frequency spectrum modelling using GMM to capture the underlying statistical information to improve the performance of face recognition system. Local spectrum features are extracted using overlap sub block window that are mapping on the face image. For each of this block, spatial domain is transformed to frequency domain using DFT. A low frequency coefficient is preserved by discarding high frequency coefficients by applying rectangular mask on the spectrum of the facial image. Low frequency information is non Gaussian in the feature space and by using combination of several Gaussian function that has different statistical properties, the best feature representation can be model using probability density function. The recognition process is performed using maximum likelihood value computed using pre-calculate GMM components. The method is tested using FERET data sets and is able to achieved 92% recognition rates.

Keywords: local features modelling, face recognition system, Gaussian mixture models, Feret

Procedia PDF Downloads 632

Authors: Tomasz Sapiński, Dorota Kamińska, Adam Pelikant, Egils Avots, Cagri Ozcinar, Gholamreza Anbarjafari

Abstract:

People express emotions through different modalities. Integration of verbal and non-verbal communication channels creates a system in which the message is easier to understand. Expanding the focus to several expression forms can facilitate research on emotion recognition as well as human-machine interaction. In this article, the authors present a Polish emotional database composed of three modalities: facial expressions, body movement and gestures, and speech. The corpora contains recordings registered in studio conditions, acted out by 16 professional actors (8 male and 8 female). The data is labeled with six basic emotions categories, according to Ekman’s emotion categories. To check the quality of performance, all recordings are evaluated by experts and volunteers. The database is available to academic community and might be useful in the study on audio-visual emotion recognition.

Keywords: body movement, emotion recognition, emotional corpus, facial expressions, gestures, multimodal database, speech

Procedia PDF Downloads 322

Authors: Krishna Mohan Bathula, Fatou Bintou Loucoubar, FNU Kaleemunnisa, Christelle Scharff, Mark Anthony De Castro

Abstract:

Voice recognition algorithms such as automatic speech recognition and text-to-speech systems with African languages can play an important role in bridging the digital divide of Artificial Intelligence in Africa, contributing to the establishment of a fully inclusive information society. This paper proposes a Deep Learning model that can classify the user responses as inputs for an interactive voice response system. A dataset with Wolof language words ‘yes’ and ‘no’ is collected as audio recordings. A two stage Data Augmentation approach is adopted for enhancing the dataset size required by the deep neural network. Data preprocessing and feature engineering with Mel-Frequency Cepstral Coefficients are implemented. Convolutional Neural Networks (CNNs) have proven to be very powerful in image classification and are promising for audio processing when sounds are transformed into spectra. For performing voice response classification, the recordings are transformed into sound frequency feature spectra and then applied image classification methodology using a deep CNN model. The inference model of this trained and reusable Wolof voice response recognition system can be integrated with many applications associated with both web and mobile platforms.

Keywords: automatic speech recognition, interactive voice response, voice response recognition, wolof word classification

Procedia PDF Downloads 86

Authors: Shaode Yu, Jiajian Meng, Bing Zhu, Hang Yu, Qiurui Sun

Abstract:

Speech emotion recognition (SER) is for recognizing human subjective emotions through audio data in-depth analysis. From speech audios, how to comprehensively extract emotional information and how to effectively fuse extracted features remain challenging. This paper presents a dual-stream SER framework that embraces both full training and transfer learning of different networks for thorough feature encoding. Besides, a plug-and-play cross-attention fusion (CAF) module is implemented for the valid integration of the dual-stream encoder output. The effectiveness of the proposed CAF module is compared to the other three fusion modules (feature summation, feature concatenation, and feature-wise linear modulation) on two databases (RAVDESS and IEMO-CAP) using different dual-stream encoders (full training network, DPCNN or TextRCNN; transfer learning network, HuBERT or Wav2Vec2). Experimental results suggest that the CAF module can effectively reconcile conflicts between features from different encoders and outperform the other three feature fusion modules on the SER task. In the future, the plug-and-play CAF module can be extended for multi-branch feature fusion, and the dual-stream SER framework can be widened for multi-stream data representation to improve the recognition performance and generalization capacity.

Keywords: speech emotion recognition, cross-attention fusion, dual-stream, pre-trained

Procedia PDF Downloads 41

Authors: Rahul Mohite, Lahcen Ouarbya

Abstract:

This paper focuses on the interplay between humans and computer systems and the ability of these systems to understand and respond to human emotions, including non-verbal communication. Current emotion recognition systems are based solely on either facial or verbal expressions. The limitation of these systems is that it requires large training data sets. The paper proposes a system for recognizing human emotions that combines both speech and emotion recognition. The system utilizes advanced techniques such as deep learning and image recognition to identify facial expressions and comprehend emotions. The results show that the proposed system, based on the combination of facial expression and speech, outperforms existing ones, which are based solely either on facial or verbal expressions. The proposed system detects human emotion with an accuracy of 86%, whereas the existing systems have an accuracy of 70% using verbal expression only and 76% using facial expression only. In this paper, the increasing significance and demand for facial recognition technology in emotion recognition are also discussed.

Keywords: facial reputation, expression reputation, deep gaining knowledge of, photo reputation, facial technology, sign processing, photo type

Procedia PDF Downloads 78

Authors: Pogula Rakesh, T. Kishore Kumar

Abstract:

Speech enhancement is a long standing problem with numerous applications like teleconferencing, VoIP, hearing aids, and speech recognition. The motivation behind this research work is to obtain a clean speech signal of higher quality by applying the optimal noise cancellation technique. Real-time adaptive filtering algorithms seem to be the best candidate among all categories of the speech enhancement methods. In this paper, we propose a speech enhancement method based on Recursive Least Squares (RLS) adaptive filter of speech signals. Experiments were performed on noisy data which was prepared by adding AWGN, Babble and Pink noise to clean speech samples at -5dB, 0dB, 5dB, and 10dB SNR levels. We then compare the noise cancellation performance of proposed RLS algorithm with existing NLMS algorithm in terms of Mean Squared Error (MSE), Signal to Noise ratio (SNR), and SNR loss. Based on the performance evaluation, the proposed RLS algorithm was found to be a better optimal noise cancellation technique for speech signals.

Keywords: adaptive filter, adaptive noise canceller, mean squared error, noise reduction, NLMS, RLS, SNR, SNR loss

Procedia PDF Downloads 448

Authors: Monica Gonzalez Machorro

Abstract:

Dementia is hard to diagnose because of the lack of early physical symptoms. Early dementia recognition is key to improving the living condition of patients. Speech technology is considered a valuable biomarker for this challenge. Recent works have utilized conventional acoustic features and machine learning methods to detect dementia in speech. BERT-like classifiers have reported the most promising performance. One constraint, nonetheless, is that these studies are either based on human transcripts or on transcripts produced by automatic speech recognition (ASR) systems. This research contribution is to explore a method that does not require transcriptions to detect early Alzheimer’s disease (AD) and mild cognitive impairment (MCI). This is achieved by fine-tuning a pre-trained ASR model for the downstream early AD and MCI tasks. To do so, a subset of the thoroughly studied Pitt Corpus is customized. The subset is balanced for class, age, and gender. Data processing also involves cropping the samples into 10-second segments. For comparison purposes, a baseline model is defined by training and testing a Random Forest with 20 extracted acoustic features using the librosa library implemented in Python. These are: zero-crossing rate, MFCCs, spectral bandwidth, spectral centroid, root mean square, and short-time Fourier transform. The baseline model achieved a 58% accuracy. To fine-tune HuBERT as a classifier, an average pooling strategy is employed to merge the 3D representations from audio into 2D representations, and a linear layer is added. The pre-trained model used is ‘hubert-large-ls960-ft’. Empirically, the number of epochs selected is 5, and the batch size defined is 1. Experiments show that our proposed method reaches a 69% balanced accuracy. This suggests that the linguistic and speech information encoded in the self-supervised ASR-based model is able to learn acoustic cues of AD and MCI.

Keywords: automatic speech recognition, early Alzheimer’s recognition, mild cognitive impairment, speech impairment

Procedia PDF Downloads 98

Authors: Thiago Spilborghs Bueno Meyer, Plinio Thomaz Aquino Junior

Abstract:

Through speech, which privileges the functional and interactive nature of the text, it is possible to ascertain the spatiotemporal circumstances, the conditions of production and reception of the discourse, the explicit purposes such as informing, explaining, convincing, etc. These conditions allow bringing the interaction between humans closer to the human-robot interaction, making it natural and sensitive to information. However, it is not enough to understand what is said; it is necessary to recognize emotions for the desired interaction. The validity of the use of neural networks for feature selection and emotion recognition was verified. For this purpose, it is proposed the use of neural networks and comparison of models, such as recurrent neural networks and deep neural networks, in order to carry out the classification of emotions through speech signals to verify the quality of recognition. It is expected to enable the implementation of robots in a domestic environment, such as the HERA robot from the RoboFEI@Home team, which focuses on autonomous service robots for the domestic environment. Tests were performed using only the Mel-Frequency Cepstral Coefficients, as well as tests with several characteristics of Delta-MFCC, spectral contrast, and the Mel spectrogram. To carry out the training, validation and testing of the neural networks, the eNTERFACE’05 database was used, which has 42 speakers from 14 different nationalities speaking the English language. The data from the chosen database are videos that, for use in neural networks, were converted into audios. It was found as a result, a classification of 51,969% of correct answers when using the deep neural network, when the use of the recurrent neural network was verified, with the classification with accuracy equal to 44.09%. The results are more accurate when only the Mel-Frequency Cepstral Coefficients are used for the classification, using the classifier with the deep neural network, and in only one case, it is possible to observe a greater accuracy by the recurrent neural network, which occurs in the use of various features and setting 73 for batch size and 100 training epochs.

Keywords: emotion recognition, speech, deep learning, human-robot interaction, neural networks

Procedia PDF Downloads 131

Authors: B. R. Campomanes-Alvarez, P. Quiros, B. Fernandez

Abstract:

Automatic Speech Recognition (ASR) is a machine-based process of decoding and transcribing oral speech. A typical ASR system receives acoustic input from a speaker or an audio file, analyzes it using algorithms, and produces an output in the form of a text. Some speech recognition systems use Hidden Markov Models (HMMs) to deal with the temporal variability of speech and Gaussian Mixture Models (GMMs) to determine how well each state of each HMM fits a short window of frames of coefficients that represents the acoustic input. Another way to evaluate the fit is to use a feed-forward neural network that takes several frames of coefficients as input and produces posterior probabilities over HMM states as output. Deep neural networks (DNNs) that have many hidden layers and are trained using new methods have been shown to outperform GMMs on a variety of speech recognition systems. Acoustic models for state-of-the-art ASR systems are usually training on massive amounts of data. However, audio files with their corresponding transcriptions can be difficult to obtain, especially in the Spanish language. Hence, in the case of these low-resource scenarios, building an ASR model is considered as a complex task due to the lack of labeled data, resulting in an under-trained system. Semi-supervised learning approaches arise as necessary tasks given the high cost of transcribing audio data. The main goal of this proposal is to develop a procedure based on acoustic semi-supervised learning for Spanish ASR systems by using DNNs. This semi-supervised learning approach consists of: (a) Training a seed ASR model with a DNN using a set of audios and their respective transcriptions. A DNN with a one-hidden-layer network was initialized; increasing the number of hidden layers in training, to a five. A refinement, which consisted of the weight matrix plus bias term and a Stochastic Gradient Descent (SGD) training were also performed. The objective function was the cross-entropy criterion. (b) Decoding/testing a set of unlabeled data with the obtained seed model. (c) Selecting a suitable subset of the validated data to retrain the seed model, thereby improving its performance on the target test set. To choose the most precise transcriptions, three confidence scores or metrics, regarding the lattice concept (based on the graph cost, the acoustic cost and a combination of both), was performed as selection technique. The performance of the ASR system will be calculated by means of the Word Error Rate (WER). The test dataset was renewed in order to extract the new transcriptions added to the training dataset. Some experiments were carried out in order to select the best ASR results. A comparison between a GMM-based model without retraining and the DNN proposed system was also made under the same conditions. Results showed that the semi-supervised ASR-model based on DNNs outperformed the GMM-model, in terms of WER, in all tested cases. The best result obtained an improvement of 6% relative WER. Hence, these promising results suggest that the proposed technique could be suitable for building ASR models in low-resource environments.

Keywords: automatic speech recognition, deep neural networks, machine learning, semi-supervised learning

Procedia PDF Downloads 318

Authors: A. Belmeguenai, K. Mansouri, R. Djemili

Abstract:

This work present a new algorithm based on the nonlinear filter generator for speech encryption and decryption. The proposed algorithm consists on the use a linear feedback shift register (LFSR) whose polynomial is primitive and nonlinear Boolean function. The purpose of this system is to construct Keystream with good statistical properties, but also easily computable on a machine with limited capacity calculated. This proposed speech encryption scheme is very simple, highly efficient, and fast to implement the speech encryption and decryption. We conclude the paper by showing that this system can resist certain known attacks.

Keywords: nonlinear filter generator, stream ciphers, speech encryption, security analysis

Procedia PDF Downloads 266

Authors: Manjare Chandraprabha A., S. D. Shirbahadurkar, Manjare Anil S., Paithne Ajay N.

Abstract:

This paper proposes Intonation Modeling for Prosody generation in Neutral speech for Marathi (language spoken in Maharashtra, India) story telling applications. Nowadays audio story telling devices are very eminent for children. In this paper, we proposed tilt model for stressed words in Marathi for speech modification. Tilt model predicts modification in tone of neutral speech. GMM is used to identify stressed words for modification.

Keywords: tilt model, fundamental frequency, statistical parametric speech synthesis, GMM

Procedia PDF Downloads 359

Authors: Sanja Seljan, Ivan Dunđer

Abstract:

The paper presents combined automatic speech recognition (ASR) for English and machine translation (MT) for English and Croatian in the domain of business correspondence. The first part presents results of training the ASR commercial system on two English data sets, enriched by error analysis. The second part presents results of machine translation performed by online tool Google Translate for English and Croatian and Croatian-English language pairs. Human evaluation in terms of usability is conducted and internal consistency calculated by Cronbach's alpha coefficient, enriched by error analysis. Automatic evaluation is performed by WER (Word Error Rate) and PER (Position-independent word Error Rate) metrics, followed by investigation of Pearson’s correlation with human evaluation.

Keywords: automatic machine translation, integrated language technologies, quality evaluation, speech recognition

Procedia PDF Downloads 453

Authors: Prasad Kalagura, B. Veeresh kumar

Abstract:

We constructed an interface system that would allow a similarly paralyzed user to interact with a computer with almost full functional capability. A real-time tracking algorithm is implemented based on adaptive skin detection and motion analysis. The clicking of the mouse is activated by the user's eye blinking through a sensor. The keyboard function is implemented by voice recognition kit.

Keywords: embedded ARM7 processor, mouse pointer control, voice recognition

Procedia PDF Downloads 549

Authors: Babatunde Olumide Olawale, Oyebode Olumide Oyediran

Abstract:

The security safe is a place or building where classified document and precious items are kept. To prevent unauthorised persons from gaining access to this safe a lot of technologies had been used. But frequent reports of an unauthorised person gaining access into security safes with the aim of removing document and items from the safes are pointers to the fact that there is still security gap in the recent technologies used as access control for the security safe. In this paper we try to solve this problem by developing a multimodal biometric system for physical access control into a security safe using face and voice recognition. The safe is accessed by the combination of face and speech pattern recognition and also in that sequential order. User authentication is achieved through the use of camera/sensor unit and a microphone unit both attached to the door of the safe. The user face was captured by the camera/sensor while the speech was captured by the use of the microphone unit. The Scale Invariance Feature Transform (SIFT) algorithm was used to train images to form templates for the face recognition system while the Mel-Frequency Cepitral Coefficients (MFCC) algorithm was used to train the speech recognition system to recognise authorise user’s speech. Both algorithms were hosted in two separate web based servers and for automatic analysis of our work; our developed system was simulated in a MATLAB environment. The results obtained shows that the developed system was able to give access to authorise users while declining unauthorised person access to the security safe.

Keywords: access control, multimodal biometrics, pattern recognition, security safe

Procedia PDF Downloads 299

Authors: Jihye Jeon

Abstract:

This paper analyzes the conceptual framework of three statistical methods, multiple regression, path analysis, and structural equation models. When establishing research model of the statistical modeling of complex social phenomenon, it is important to know the strengths and limitations of three statistical models. This study explored the character, strength, and limitation of each modeling and suggested some strategies for accurate explaining or predicting the causal relationships among variables. Especially, on the studying of depression or mental health, the common mistakes of research modeling were discussed.

Keywords: multiple regression, path analysis, structural equation models, statistical modeling, social and psychological phenomenon

Procedia PDF Downloads 602

Authors: Sergei Aleinik, Yuri Matveev

Abstract:

In this paper a novel method for the detection of clipping in speech signals is described. It is shown that the new method has better performance than known clipping detection methods, is easy to implement, and is robust to changes in signal amplitude, size of data, etc. Statistical simulation results are presented.

Keywords: clipping, clipped signal, speech signal processing, digital signal processing

Procedia PDF Downloads 363

Authors: Yishu Gong, Liangliang Yang, Jianyu Zhang, Zhengyu Chen, Sihong He, Xusheng Zhang, Wei Zhang

Abstract:

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that affects millions of people worldwide and is characterized by cognitive decline and behavioral changes. People living with Alzheimer’s disease often find it hard to complete routine tasks. However, there are limited objective assessments that aim to quantify the difficulty of certain tasks for AD patients compared to non-AD people. In this study, we propose to use speech emotion recognition (SER), especially the frustration level, as a potential biomarker for quantifying the difficulty patients experience when describing a picture. We build an SER model using data from the IEMOCAP dataset and apply the model to the DementiaBank data to detect the AD/non-AD group difference and perform longitudinal analysis to track the AD disease progression. Our results show that the frustration level detected from the SER model can possibly be used as a cost-effective tool for objective tracking of AD progression in addition to the Mini-Mental State Examination (MMSE) score.

Keywords: Alzheimer’s disease, speech emotion recognition, longitudinal biomarker, machine learning

Procedia PDF Downloads 82

Authors: Hajer Rahali, Zied Hajaiej, Noureddine Ellouze

Abstract:

The goal of speech parameterization is to extract the relevant information about what is being spoken from the audio signal. In speech recognition systems Mel-Frequency Cepstral Coefficients (MFCC) and Relative Spectral Mel-Frequency Cepstral Coefficients (RASTA-MFCC) are the two main techniques used. It will be shown in this paper that it presents some modifications to the original MFCC method. In our work the effectiveness of proposed changes to MFCC called Modified Function Cepstral Coefficients (MODFCC) were tested and compared against the original MFCC and RASTA-MFCC features. The prosodic features such as jitter and shimmer are added to baseline spectral features. The above-mentioned techniques were tested with impulsive signals under various noisy conditions within AURORA databases.

Keywords: auditory filter, impulsive noise, MFCC, prosodic features, RASTA filter

Procedia PDF Downloads 396

Authors: Albatool Al Abdulwahid, Bayan Shakally, Mariam Mohamed, Wed Almokri

Abstract:

Artificial intelligence has infiltrated every part of our life and every field we can think of. With technical developments, artificial intelligence applications are becoming more prevalent. We chose ELSA speak because it is a magnificent example of Artificial intelligent applications, ELSA speak is a smartphone application that is free to download on both IOS and Android smartphones. ELSA speak utilizes artificial intelligence to help non-native English speakers pronounce words and phrases similar to a native speaker, as well as enhance their English skills. It employs speech-recognition technology that aids the application to excel the pronunciation of its users. This remarkable feature distinguishes ELSA from other voice recognition algorithms and increase the efficiency of the application. This study focused on evaluating ELSA speak application, by testing the degree of effectiveness based on survey questions. The results of the questionnaire were variable. The generality of the participants strongly agreed that ELSA has helped them enhance their pronunciation skills. However, a few participants were unconfident about the application’s ability to assist them in their learning journey.

Keywords: ELSA speak application, artificial intelligence, speech-recognition technology, language learning, english pronunciation

Procedia PDF Downloads 73