Federated Averaging (FedAvg) is a leading decentralized machine learning approach, prioritizing d... more Federated Averaging (FedAvg) is a leading decentralized machine learning approach, prioritizing data privacy. However, it faces challenges like non-identically distributed data, communication bottlenecks, and adversarial attacks. This abstract introduces a fuzzy-based FedAvg, leveraging fuzzy logic to manage uncertainty in decentralized environments. Fuzzy clustering adapts the model to varied data distributions, addressing non-IID challenges. Fuzzy membership functions enhance aggregation by introducing an adaptive weighting scheme, improving convergence and accuracy. The fuzzy approach incorporates privacy-preserving mechanisms, ensuring secure aggregation with homomorphic encryption and differential privacy. Simulations show improved convergence, resilience to non-IID data, and enhanced privacy compared to traditional FedAvg, contributing to more secure decentralized ML systems.
The potential uses of emotion recognition in domains like human-robot interaction, marketing, emo... more The potential uses of emotion recognition in domains like human-robot interaction, marketing, emotional gaming, and humancomputer interface have made it a prominent research subject. Better user experiences can result from the development of technologies that can accurately interpret and respond to human emotions thanks to a better understanding of emotions. The use of several sensors and computational algorithms is the main emphasis of this paper's thorough analysis of the developments in emotion recognition techniques. Our results show that using more than one modality improves the performance of emotion recognition when a variety of metrics and computational techniques are used. This paper adds to the body of knowledge by thoroughly examining and contrasting several state-of-art computational techniques and measurements for emotion recognition. The study emphasizes how crucial it is to use a variety of modalities along with cutting-edge machine learning algorithms in order to attain more precise and trustworthy emotion assessment. Additionally, we pinpoint prospective avenues for additional investigation and advancement, including the incorporation of multimodal data and the investigation of innovative features and fusion methodologies. This study contributes to the development of technology that can better comprehend and react to human emotions by offering practitioners and academics in the field of emotion recognition insightful advice.
Labeling speech in an audio file with appropriate dialect labels is the aim of a dialect identifi... more Labeling speech in an audio file with appropriate dialect labels is the aim of a dialect identification system. This paper presents a method of using convolution neural networks (CNN) to identify four Assamese dialects: Goalporia dialect, Kamrupi dialect, Eastern Assamese dialect, and Central Assamese dialect. This study employed the speech patterns of four major Assamese regional dialects: the Central Dialects spoken in and around the district of Nagaon; the Eastern Assamese dialect spoken in the districts of Sibsagar and its neighboring areas; the Kamrupi dialect spoken in the districts of Kamrup, Nalbari, Barpeta, Kokarajhar, and some areas of Bongaigaon; and the Goaplari dialect spoken in the Goaplara, Dhuburi, and a portion of Bongaigaon district. Over the course of two hours, audio samples from each of the four dialects were used to train the classifier. Mel spectrogram pictures, which are produced from two to four second divisions of raw audio input with varying audio quality, are used by the CNN. The system's performance is also analyzed in relation to the lengths of the train and test audio samples. The proposed CNN model achieves an accuracy of 90.82 percent, which may be the best when compared to machine learning models.
Heart disease stands as one of the most prevalent causes of death globally, and India is no excep... more Heart disease stands as one of the most prevalent causes of death globally, and India is no exception. In India, a considerable number of fatalities can be attributed to heart-related issues each year. Factors like high blood pressure, diabetes, smoking habits, and increasingly inactive lifestyles have further contributed to the growing cases of heart disease in the country. Identifying the disease at an early stage, as well as predicting its likelihood, plays a crucial role in enhancing patient outcomes and minimizing the overall healthcare burden.This research paper explores the development and evaluation of predictive models for heart disease using a dataset comprising various clinical and demographic features. Leveraging machine learning techniques, including Convolutional Neural Networks (CNN) and other models, we aim to identify key predictors of heart disease and develop an accurate model to assist healthcare professionals in early diagnosis and intervention. The study employs comprehensive data preprocessing, feature selection, and model evaluation to ensure robust and reliable predictions. Our findings highlight the potential of machine learning and deep learning models to significantly enhance heart disease prediction, thereby contributing to better management and prevention strategies in India.
Speaker diarization is a task to identify the speaker when different speakers spoke in an audio o... more Speaker diarization is a task to identify the speaker when different speakers spoke in an audio or video recording environment. Artificial intelligence (AI) fields have effectively used Deep Learning (DL) to solve a variety of real-world application challenges. With effective applications in a wide range of subdomains, such as natural language processing, image processing, computer vision, speech and speaker recognition, and emotion recognition, cyber security, and many others, DL, a very innovative field of Machine Learning (ML), that is quickly emerging as the most potent machine learning technique. DL techniques have outperformed conventional approaches recently in speaker diarization as well as speaker recognition. The technique of assigning classes to speech recordings that correspond to the speaker's identity is known as speaker diarization, and it allows one to determine who spoke when. A crucial step in speech processing is speaker diarization, which divides an audio recording into different speaker areas. In-depth analysis of speaker diarization utilizing a variety of deep learning algorithms that are presented in this research paper. NIST-2000 CALLHOME and our in-house database ALSD-DB are the two voice corpora we used for this study's tests. TDNN-based embeddings with x-vectors, LSTM-based embeddings with d-vectors, and lastly embeddings fusion of both x-vector and d-vector are used in the tests for the basic system. For the NIST-2000 CALLHOME database, LSTM based embeddings with d-vector and embeddings integrating both x-vector and d-vector exhibit improved performance with DER of 8.25% and 7.65%, respectively, and of 10.45% and 9.65% for the local ALSD-DB database.
In sophisticated Human-Computer Interfaces (HCI), the emotional state of the user is becoming a c... more In sophisticated Human-Computer Interfaces (HCI), the emotional state of the user is becoming a crucial component that is closely linked to emotional speech recognition. Spoken expressions, which can be a part of human-machine interaction, are an important source of emotional information. Speech emotion recognition (SER) in deep learning (DL) continues to be a hot topic, especially in the field of emotional computing. Current deep learning (DL) and neural network methods are applied in this highly active field of research. This is as a result of its expanding potential, advancements in algorithms, and practical uses. Quantitative factors such as pitch, intensity, accent and Mel-Frequency Cepstral Coefficients (MFCC) can be employed to model the paralinguistic data contained in human speech. To achieve SER, three key procedures are usually followed: data processing, feature selection/extraction, and classification based on the underlying emotional qualities. The nature of these processes and the peculiarities of human speech lend support to the employment of DL techniques for SER implementation. A variety of DL methods have been used for SER tasks in recent affective computing research works; however, only a small number of them capture the underlying ideas and methodologies that can be used to facilitate the three main steps of SER implementation. With a focus on the three SER implementation processes, we provide a state-of-the-art assessment of research conducted over the last ten years that tackled SER tasks from DL perspectives in this work. Various issues are covered in detail, including the problem of low classification accuracy of Speaker-Independent experiments and the related remedies. The review offers principles for SER evaluation as well, emphasizing indicators that can be experimented with and common baselines.
International Journal of Scientific Research in Computer Science and Engineering, 2017
In this paper a brief comparison studies on the performance of different speaker modeling techniq... more In this paper a brief comparison studies on the performance of different speaker modeling techniques in robust and reliable speaker verification (SV) system has been discussed. In text-independent speaker verification, lots of states of art speaker modeling techniques have been developed in different scenarios to upgrade its performance. The performance of SV system is not only depended on the fusion of different feature vectors but also it is highly depended upon the fusion of various speaker modeling techniques. In this work, an automatic SV system has been developed using the Mel-Frequency Cepstral Coefficients (MFCC) combined with the Prosodic feature vectors. The baseline of the SV system has been trained with speaker modeling techniques separately and fusions namely Vector Quantization (VQ), Gaussian Mixture Model (GMM), GMM-Universal Background Model (GMM-UBM), Support Vector Machine (SVM) and Joint Factor Analysis (JFA) to analyze its performances. The results reported here, have been evaluated using the multilingual speech database, namely Arunachali Language Speech Database (ALS-DB). From the experimental point of view we observe that the best performance of SV system shows by JFA with GMM-UBM modeling technique with its EER value of 4.76% and MinDCF value of 0.0872. Comparing with other modeling techniques VQ shows its poor performance with its EER value of 11.08% and MinDCF value of 0.2010. SVM shows of approximately 2.8% improvement of verification rate with comparison to that of GMM-UBM. Here, finally, we conclude that the fusions of both generative and discriminative models highly improve the performance of SV system.
International journal of membrane science and technology, Jun 22, 2023
Labeling speech in an audio file with appropriate dialect labels is the aim of a dialect identifi... more Labeling speech in an audio file with appropriate dialect labels is the aim of a dialect identification system. This paper presents a method of using convolution neural networks (CNN) to identify four Assamese dialects: Goalporia dialect, Kamrupi dialect, Eastern Assamese dialect, and Central Assamese dialect. This study employed the speech patterns of four major Assamese regional dialects: the Central Dialects spoken in and around the district of Nagaon; the Eastern Assamese dialect spoken in the districts of Sibsagar and its neighboring areas; the Kamrupi dialect spoken in the districts of Kamrup, Nalbari, Barpeta, Kokarajhar, and some areas of Bongaigaon; and the Goaplari dialect spoken in the Goaplara, Dhuburi, and a portion of Bongaigaon district. Over the course of two hours, audio samples from each of the four dialects were used to train the classifier. Mel spectrogram pictures, which are produced from two to four second divisions of raw audio input with varying audio quality, are used by the CNN. The system's performance is also analyzed in relation to the lengths of the train and test audio samples. The proposed CNN model achieves an accuracy of 90.82 percent, which may be the best when compared to machine learning models.
The performance of common speaker verification (SV) system vastly affected when speaker model tra... more The performance of common speaker verification (SV) system vastly affected when speaker model training is done in the speech sample that recorded done by one device and the testing is done in another device. It is a major problem of speaker verification system in multi-device environment. In this paper we report the experiment carried out on the recently collected speaker recognition database Arunachali Language Speech Database (ALS-DB) a multilingual and multichannel database to study the impact of device variability on speaker verification system. The collected database is evaluated with Gaussian mixture model and Universal Background Model (GMM-UBM) and Mel - Frequency Cepstral Coefficients (MFCC) combined with prosodic features as a front end feature vectors based speaker verification system. The impact of the device both matching and mismatch in training and testing has been evaluated in text independent manner. For matching condition of device we have found Equal Error Rate (EER) 7.50% with minimum Detection Cost Function (MinDCF) value 0.1062 and for mismatching condition of devices that of 18.70% with MinDCF value 0.3425. The performance of the SV system has degraded approximately 11.00% due to mismatching condition of devices in text independent speaker verification system.
International journal of computer applications, Jan 16, 2014
Language Identification (LID) is one of the most popular areas of research in speech signal proce... more Language Identification (LID) is one of the most popular areas of research in speech signal processing. Now a day's lots of approaches have been used to improve performance of LID system which includes Parallel Phone Recognition Language Modeling (PPRLM), Support Vector Machine (SVM) and general Gaussian Mixture Model (GMM) etc. The state-of-art LID system has been utilised lots of feature vectors like LPCC, MFCC, SDC and prosodic. Although fusion of prosodic features with MFCC features shows some improvement in the performance of the LID system. But still it is not sufficient. In this paper, a baseline system for the LID system in multilingual environments has been developed using GMM as a classifier and MFCC combined with Shifted-Delta-Cepstral (SDC) as front end processing feature vectors. In this works, we used the Arunachali Language Speech Database (ALS-DB), a multilingual and multichannel speech corpus which was recently collected from the four local languages namely Adi, Apatani, Galo and Nyishi in Arunachal Pradesh including Hindi and English as secondary languages.The performance of the LID system has been improved by combing MFCC and SDC features than its individual performances. The minimum ERR rates for the features MFCC and SDC individually are 19.70% and 11.83% respectively while minimum ERR rate for the combined features both MFCC and SDC is 6.40%.Approximately 15.00% and 6.00% of performance of the LID system has been improved while using the combining features of MFCC with SDC over the baseline systems that using MFCC and SDC features in individual respectively.
2013 International Conference on Intelligent Systems and Signal Processing (ISSP), 2013
This paper report the experiments carried out on a recently collected multilingual speech databas... more This paper report the experiments carried out on a recently collected multilingual speech database namely Arunachali Language Speech Database (ALS-DB) to identify the spoken language of the speaker. The speech database consists of speech data recorded from 200 speakers with Arunachali languages of North-East India as mother tongue. The speech data is collected in three different languages English, Hindi and a local language which belongs to any one of the four major languages of Arunachal Pradesh: Adi, Nyishi, Galo and Apatani. The collected database is evaluated with Gaussian mixture model (GMM) based language identification system with MFCC and MFCC with Prosodic features as feature vector. The initial study explores the fact that performance of a baseline GMM-MFCC based language identification system improves considerably when the prosodic features are considered as additional features with MFCC. It has been observed that when prosodic features are combined with MFCC features, performance of the system improved by nearly 11% over the baseline performance.
— Automatic Speaker Verification (ASV) refers to the task of verifying the claimed identity of a ... more — Automatic Speaker Verification (ASV) refers to the task of verifying the claimed identity of a speaker based on speech data. The decision made by a Speaker Verification system is basically a binary decision returns either ―Yes ‖ or ―No ‖ based on the credibility of the claim, determined by some scoring techniques. The output of an automatic speaker verification system is highly dependent on database used for training and testing the system. The results obtained by the speaker verification system are meaningless if recording specifications and environment for training and testing data are not known. This paper describes methodology and experimental setup used for the development of a speech corpus for the evaluation of textindependent speaker verification system in multilingual environment. Four major languages of Arunachal Pradesh (a North-Eastern frontier state of India, boarding with China) Nyishi, Adi, Galo and Apatani along with English and Hindi have been considered for the d...
Automatic Speaker Verification (ASV) refers to the task of verifying the claimed identity of a sp... more Automatic Speaker Verification (ASV) refers to the task of verifying the claimed identity of a speaker based on speech data. The decision made by a Speaker Verification system is basically a binary decision returns either ―Yes‖ or ―No‖ based on the credibility of the claim, determined by some scoring techniques. The output of an automatic speaker verification system is highly dependent on database used for training and testing the system. The results obtained by the speaker verification system are meaningless if recording specifications and environment for training and testing data are not known. This paper describes methodology and experimental setup used for the development of a speech corpus for the evaluation of text- independent speaker verification system in multilingual environment. Four major languages of Arunachal Pradesh (a North-Eastern frontier state of India, boarding with China) Nyishi, Adi, Galo and Apatani along with English and Hindi have been considered for the dev...
The performance of common speaker verification (SV) system vastly affected when speaker model tra... more The performance of common speaker verification (SV) system vastly affected when speaker model training is done in the speech sample that recorded done by one device and the testing is done in another device. It is a major problem of speaker verification system in multi-device environment. In this paper we report the experiment carried out on the recently collected speaker recognition database Arunachali Language Speech Database (ALS-DB) a multilingual and multichannel database to study the impact of device variability on speaker verification system. The collected database is evaluated with Gaussian mixture model and Universal Background Model (GMM-UBM) and Mel - Frequency Cepstral Coefficients (MFCC) combined with prosodic features as a front end feature vectors based speaker verification system. The impact of the device both matching and mismatch in training and testing has been evaluated in text independent manner. For matching condition of device we have found Equal Error Rate (E...
Speaker modeling distance normalization (D-Norm) is one of the important score normalization tech... more Speaker modeling distance normalization (D-Norm) is one of the important score normalization techniques in speaker verification (SV) system. For D-Norm implementation, it doesn’t need any additional speech data or external speaker population, that being one of the essential advantages of D-Norm. But traditional D-Norm has still disadvantage of its time complexity due to the Monte-Carlo based Kullback –Leibler distance estimation approach. In this paper we used a simplified an innovative D-Norm technique, which is based on the upper bound of the KL divergence between two statistical models namely GMM-UBM adapted target speaker model and the UBM model. This approach proved better performance. From the experimental point of view we can conclude that the performance of the multilingual SV system has been degrades approximately 3.0% due to language mismatching of the training and testing of the system. Also by applying D-Norm, the performance of the baseline system has been improved appr...
Federated Averaging (FedAvg) is a leading decentralized machine learning approach, prioritizing d... more Federated Averaging (FedAvg) is a leading decentralized machine learning approach, prioritizing data privacy. However, it faces challenges like non-identically distributed data, communication bottlenecks, and adversarial attacks. This abstract introduces a fuzzy-based FedAvg, leveraging fuzzy logic to manage uncertainty in decentralized environments. Fuzzy clustering adapts the model to varied data distributions, addressing non-IID challenges. Fuzzy membership functions enhance aggregation by introducing an adaptive weighting scheme, improving convergence and accuracy. The fuzzy approach incorporates privacy-preserving mechanisms, ensuring secure aggregation with homomorphic encryption and differential privacy. Simulations show improved convergence, resilience to non-IID data, and enhanced privacy compared to traditional FedAvg, contributing to more secure decentralized ML systems.
The potential uses of emotion recognition in domains like human-robot interaction, marketing, emo... more The potential uses of emotion recognition in domains like human-robot interaction, marketing, emotional gaming, and humancomputer interface have made it a prominent research subject. Better user experiences can result from the development of technologies that can accurately interpret and respond to human emotions thanks to a better understanding of emotions. The use of several sensors and computational algorithms is the main emphasis of this paper's thorough analysis of the developments in emotion recognition techniques. Our results show that using more than one modality improves the performance of emotion recognition when a variety of metrics and computational techniques are used. This paper adds to the body of knowledge by thoroughly examining and contrasting several state-of-art computational techniques and measurements for emotion recognition. The study emphasizes how crucial it is to use a variety of modalities along with cutting-edge machine learning algorithms in order to attain more precise and trustworthy emotion assessment. Additionally, we pinpoint prospective avenues for additional investigation and advancement, including the incorporation of multimodal data and the investigation of innovative features and fusion methodologies. This study contributes to the development of technology that can better comprehend and react to human emotions by offering practitioners and academics in the field of emotion recognition insightful advice.
Labeling speech in an audio file with appropriate dialect labels is the aim of a dialect identifi... more Labeling speech in an audio file with appropriate dialect labels is the aim of a dialect identification system. This paper presents a method of using convolution neural networks (CNN) to identify four Assamese dialects: Goalporia dialect, Kamrupi dialect, Eastern Assamese dialect, and Central Assamese dialect. This study employed the speech patterns of four major Assamese regional dialects: the Central Dialects spoken in and around the district of Nagaon; the Eastern Assamese dialect spoken in the districts of Sibsagar and its neighboring areas; the Kamrupi dialect spoken in the districts of Kamrup, Nalbari, Barpeta, Kokarajhar, and some areas of Bongaigaon; and the Goaplari dialect spoken in the Goaplara, Dhuburi, and a portion of Bongaigaon district. Over the course of two hours, audio samples from each of the four dialects were used to train the classifier. Mel spectrogram pictures, which are produced from two to four second divisions of raw audio input with varying audio quality, are used by the CNN. The system's performance is also analyzed in relation to the lengths of the train and test audio samples. The proposed CNN model achieves an accuracy of 90.82 percent, which may be the best when compared to machine learning models.
Heart disease stands as one of the most prevalent causes of death globally, and India is no excep... more Heart disease stands as one of the most prevalent causes of death globally, and India is no exception. In India, a considerable number of fatalities can be attributed to heart-related issues each year. Factors like high blood pressure, diabetes, smoking habits, and increasingly inactive lifestyles have further contributed to the growing cases of heart disease in the country. Identifying the disease at an early stage, as well as predicting its likelihood, plays a crucial role in enhancing patient outcomes and minimizing the overall healthcare burden.This research paper explores the development and evaluation of predictive models for heart disease using a dataset comprising various clinical and demographic features. Leveraging machine learning techniques, including Convolutional Neural Networks (CNN) and other models, we aim to identify key predictors of heart disease and develop an accurate model to assist healthcare professionals in early diagnosis and intervention. The study employs comprehensive data preprocessing, feature selection, and model evaluation to ensure robust and reliable predictions. Our findings highlight the potential of machine learning and deep learning models to significantly enhance heart disease prediction, thereby contributing to better management and prevention strategies in India.
Speaker diarization is a task to identify the speaker when different speakers spoke in an audio o... more Speaker diarization is a task to identify the speaker when different speakers spoke in an audio or video recording environment. Artificial intelligence (AI) fields have effectively used Deep Learning (DL) to solve a variety of real-world application challenges. With effective applications in a wide range of subdomains, such as natural language processing, image processing, computer vision, speech and speaker recognition, and emotion recognition, cyber security, and many others, DL, a very innovative field of Machine Learning (ML), that is quickly emerging as the most potent machine learning technique. DL techniques have outperformed conventional approaches recently in speaker diarization as well as speaker recognition. The technique of assigning classes to speech recordings that correspond to the speaker's identity is known as speaker diarization, and it allows one to determine who spoke when. A crucial step in speech processing is speaker diarization, which divides an audio recording into different speaker areas. In-depth analysis of speaker diarization utilizing a variety of deep learning algorithms that are presented in this research paper. NIST-2000 CALLHOME and our in-house database ALSD-DB are the two voice corpora we used for this study's tests. TDNN-based embeddings with x-vectors, LSTM-based embeddings with d-vectors, and lastly embeddings fusion of both x-vector and d-vector are used in the tests for the basic system. For the NIST-2000 CALLHOME database, LSTM based embeddings with d-vector and embeddings integrating both x-vector and d-vector exhibit improved performance with DER of 8.25% and 7.65%, respectively, and of 10.45% and 9.65% for the local ALSD-DB database.
In sophisticated Human-Computer Interfaces (HCI), the emotional state of the user is becoming a c... more In sophisticated Human-Computer Interfaces (HCI), the emotional state of the user is becoming a crucial component that is closely linked to emotional speech recognition. Spoken expressions, which can be a part of human-machine interaction, are an important source of emotional information. Speech emotion recognition (SER) in deep learning (DL) continues to be a hot topic, especially in the field of emotional computing. Current deep learning (DL) and neural network methods are applied in this highly active field of research. This is as a result of its expanding potential, advancements in algorithms, and practical uses. Quantitative factors such as pitch, intensity, accent and Mel-Frequency Cepstral Coefficients (MFCC) can be employed to model the paralinguistic data contained in human speech. To achieve SER, three key procedures are usually followed: data processing, feature selection/extraction, and classification based on the underlying emotional qualities. The nature of these processes and the peculiarities of human speech lend support to the employment of DL techniques for SER implementation. A variety of DL methods have been used for SER tasks in recent affective computing research works; however, only a small number of them capture the underlying ideas and methodologies that can be used to facilitate the three main steps of SER implementation. With a focus on the three SER implementation processes, we provide a state-of-the-art assessment of research conducted over the last ten years that tackled SER tasks from DL perspectives in this work. Various issues are covered in detail, including the problem of low classification accuracy of Speaker-Independent experiments and the related remedies. The review offers principles for SER evaluation as well, emphasizing indicators that can be experimented with and common baselines.
International Journal of Scientific Research in Computer Science and Engineering, 2017
In this paper a brief comparison studies on the performance of different speaker modeling techniq... more In this paper a brief comparison studies on the performance of different speaker modeling techniques in robust and reliable speaker verification (SV) system has been discussed. In text-independent speaker verification, lots of states of art speaker modeling techniques have been developed in different scenarios to upgrade its performance. The performance of SV system is not only depended on the fusion of different feature vectors but also it is highly depended upon the fusion of various speaker modeling techniques. In this work, an automatic SV system has been developed using the Mel-Frequency Cepstral Coefficients (MFCC) combined with the Prosodic feature vectors. The baseline of the SV system has been trained with speaker modeling techniques separately and fusions namely Vector Quantization (VQ), Gaussian Mixture Model (GMM), GMM-Universal Background Model (GMM-UBM), Support Vector Machine (SVM) and Joint Factor Analysis (JFA) to analyze its performances. The results reported here, have been evaluated using the multilingual speech database, namely Arunachali Language Speech Database (ALS-DB). From the experimental point of view we observe that the best performance of SV system shows by JFA with GMM-UBM modeling technique with its EER value of 4.76% and MinDCF value of 0.0872. Comparing with other modeling techniques VQ shows its poor performance with its EER value of 11.08% and MinDCF value of 0.2010. SVM shows of approximately 2.8% improvement of verification rate with comparison to that of GMM-UBM. Here, finally, we conclude that the fusions of both generative and discriminative models highly improve the performance of SV system.
International journal of membrane science and technology, Jun 22, 2023
Labeling speech in an audio file with appropriate dialect labels is the aim of a dialect identifi... more Labeling speech in an audio file with appropriate dialect labels is the aim of a dialect identification system. This paper presents a method of using convolution neural networks (CNN) to identify four Assamese dialects: Goalporia dialect, Kamrupi dialect, Eastern Assamese dialect, and Central Assamese dialect. This study employed the speech patterns of four major Assamese regional dialects: the Central Dialects spoken in and around the district of Nagaon; the Eastern Assamese dialect spoken in the districts of Sibsagar and its neighboring areas; the Kamrupi dialect spoken in the districts of Kamrup, Nalbari, Barpeta, Kokarajhar, and some areas of Bongaigaon; and the Goaplari dialect spoken in the Goaplara, Dhuburi, and a portion of Bongaigaon district. Over the course of two hours, audio samples from each of the four dialects were used to train the classifier. Mel spectrogram pictures, which are produced from two to four second divisions of raw audio input with varying audio quality, are used by the CNN. The system's performance is also analyzed in relation to the lengths of the train and test audio samples. The proposed CNN model achieves an accuracy of 90.82 percent, which may be the best when compared to machine learning models.
The performance of common speaker verification (SV) system vastly affected when speaker model tra... more The performance of common speaker verification (SV) system vastly affected when speaker model training is done in the speech sample that recorded done by one device and the testing is done in another device. It is a major problem of speaker verification system in multi-device environment. In this paper we report the experiment carried out on the recently collected speaker recognition database Arunachali Language Speech Database (ALS-DB) a multilingual and multichannel database to study the impact of device variability on speaker verification system. The collected database is evaluated with Gaussian mixture model and Universal Background Model (GMM-UBM) and Mel - Frequency Cepstral Coefficients (MFCC) combined with prosodic features as a front end feature vectors based speaker verification system. The impact of the device both matching and mismatch in training and testing has been evaluated in text independent manner. For matching condition of device we have found Equal Error Rate (EER) 7.50% with minimum Detection Cost Function (MinDCF) value 0.1062 and for mismatching condition of devices that of 18.70% with MinDCF value 0.3425. The performance of the SV system has degraded approximately 11.00% due to mismatching condition of devices in text independent speaker verification system.
International journal of computer applications, Jan 16, 2014
Language Identification (LID) is one of the most popular areas of research in speech signal proce... more Language Identification (LID) is one of the most popular areas of research in speech signal processing. Now a day's lots of approaches have been used to improve performance of LID system which includes Parallel Phone Recognition Language Modeling (PPRLM), Support Vector Machine (SVM) and general Gaussian Mixture Model (GMM) etc. The state-of-art LID system has been utilised lots of feature vectors like LPCC, MFCC, SDC and prosodic. Although fusion of prosodic features with MFCC features shows some improvement in the performance of the LID system. But still it is not sufficient. In this paper, a baseline system for the LID system in multilingual environments has been developed using GMM as a classifier and MFCC combined with Shifted-Delta-Cepstral (SDC) as front end processing feature vectors. In this works, we used the Arunachali Language Speech Database (ALS-DB), a multilingual and multichannel speech corpus which was recently collected from the four local languages namely Adi, Apatani, Galo and Nyishi in Arunachal Pradesh including Hindi and English as secondary languages.The performance of the LID system has been improved by combing MFCC and SDC features than its individual performances. The minimum ERR rates for the features MFCC and SDC individually are 19.70% and 11.83% respectively while minimum ERR rate for the combined features both MFCC and SDC is 6.40%.Approximately 15.00% and 6.00% of performance of the LID system has been improved while using the combining features of MFCC with SDC over the baseline systems that using MFCC and SDC features in individual respectively.
2013 International Conference on Intelligent Systems and Signal Processing (ISSP), 2013
This paper report the experiments carried out on a recently collected multilingual speech databas... more This paper report the experiments carried out on a recently collected multilingual speech database namely Arunachali Language Speech Database (ALS-DB) to identify the spoken language of the speaker. The speech database consists of speech data recorded from 200 speakers with Arunachali languages of North-East India as mother tongue. The speech data is collected in three different languages English, Hindi and a local language which belongs to any one of the four major languages of Arunachal Pradesh: Adi, Nyishi, Galo and Apatani. The collected database is evaluated with Gaussian mixture model (GMM) based language identification system with MFCC and MFCC with Prosodic features as feature vector. The initial study explores the fact that performance of a baseline GMM-MFCC based language identification system improves considerably when the prosodic features are considered as additional features with MFCC. It has been observed that when prosodic features are combined with MFCC features, performance of the system improved by nearly 11% over the baseline performance.
— Automatic Speaker Verification (ASV) refers to the task of verifying the claimed identity of a ... more — Automatic Speaker Verification (ASV) refers to the task of verifying the claimed identity of a speaker based on speech data. The decision made by a Speaker Verification system is basically a binary decision returns either ―Yes ‖ or ―No ‖ based on the credibility of the claim, determined by some scoring techniques. The output of an automatic speaker verification system is highly dependent on database used for training and testing the system. The results obtained by the speaker verification system are meaningless if recording specifications and environment for training and testing data are not known. This paper describes methodology and experimental setup used for the development of a speech corpus for the evaluation of textindependent speaker verification system in multilingual environment. Four major languages of Arunachal Pradesh (a North-Eastern frontier state of India, boarding with China) Nyishi, Adi, Galo and Apatani along with English and Hindi have been considered for the d...
Automatic Speaker Verification (ASV) refers to the task of verifying the claimed identity of a sp... more Automatic Speaker Verification (ASV) refers to the task of verifying the claimed identity of a speaker based on speech data. The decision made by a Speaker Verification system is basically a binary decision returns either ―Yes‖ or ―No‖ based on the credibility of the claim, determined by some scoring techniques. The output of an automatic speaker verification system is highly dependent on database used for training and testing the system. The results obtained by the speaker verification system are meaningless if recording specifications and environment for training and testing data are not known. This paper describes methodology and experimental setup used for the development of a speech corpus for the evaluation of text- independent speaker verification system in multilingual environment. Four major languages of Arunachal Pradesh (a North-Eastern frontier state of India, boarding with China) Nyishi, Adi, Galo and Apatani along with English and Hindi have been considered for the dev...
The performance of common speaker verification (SV) system vastly affected when speaker model tra... more The performance of common speaker verification (SV) system vastly affected when speaker model training is done in the speech sample that recorded done by one device and the testing is done in another device. It is a major problem of speaker verification system in multi-device environment. In this paper we report the experiment carried out on the recently collected speaker recognition database Arunachali Language Speech Database (ALS-DB) a multilingual and multichannel database to study the impact of device variability on speaker verification system. The collected database is evaluated with Gaussian mixture model and Universal Background Model (GMM-UBM) and Mel - Frequency Cepstral Coefficients (MFCC) combined with prosodic features as a front end feature vectors based speaker verification system. The impact of the device both matching and mismatch in training and testing has been evaluated in text independent manner. For matching condition of device we have found Equal Error Rate (E...
Speaker modeling distance normalization (D-Norm) is one of the important score normalization tech... more Speaker modeling distance normalization (D-Norm) is one of the important score normalization techniques in speaker verification (SV) system. For D-Norm implementation, it doesn’t need any additional speech data or external speaker population, that being one of the essential advantages of D-Norm. But traditional D-Norm has still disadvantage of its time complexity due to the Monte-Carlo based Kullback –Leibler distance estimation approach. In this paper we used a simplified an innovative D-Norm technique, which is based on the upper bound of the KL divergence between two statistical models namely GMM-UBM adapted target speaker model and the UBM model. This approach proved better performance. From the experimental point of view we can conclude that the performance of the multilingual SV system has been degrades approximately 3.0% due to language mismatching of the training and testing of the system. Also by applying D-Norm, the performance of the baseline system has been improved appr...
Uploads
Papers by Kshirod Sarmah