Dialect Identification of Assamese Language using Spectral Features

Dialect Identification is the process of identifies the dialects of particular standard language. The Telugu Language is one of the historical and important languages. Like any other language Telugu also contains mainly three dialects Telangana, Costa Andhra and Rayalaseema. The research work in dialect identification is very less compare to Language identification because of dearth of database. In any dialects identification system, the database and feature engineering play vital roles because of most the words are similar in pronunciation and also most of the researchers apply statistical approaches like Hidden Markov Model (HMM), Gaussian Mixture Model (GMM), etc. to work on speech processing applications. But in today's world, neural networks play a vital role in all application domains and produce good results. One of the types of the neural networks is Deep Neural Networks (DNN) and it is used to achieve the state of the art performance in several fields such as speech recognition, speaker identification. In this, the Deep Neural Network (DNN) based model Multilayer Perceptron is used to identify the regional dialects of the Telugu Language using enhanced Mel Frequency Cepstral Coefficients (MFCC) features. To do this, created a database of the Telugu dialects with the duration of 5h and 45m collected from different speakers in different environments. The results produced by DNN model compared with HMM and GMM model and it is observed that the DNN model provides good performance.

This paper presents a comparison between several features normalization methods, and a comparison between different types of Gaussian Mixture Model (GMM) based supervectors normalizations for robust Speaker Verification. We implemented the methods of normalizations as a part of speaker verification system using Support Vector Machine (SVM) classifier and GMM-based supervectors. When implementing the speaker recognition system, we used Mel Frequency Cepstral Coefficients (MFCC) feature extraction. A valid question is which features normalization to use, if any. We examine the most common methods of feature normalizations, such as: Feature Warping mapping, and Cepstral Mean Subtraction (CMS) normalization with and without variance normalization. These methods were compared to features without normalization at all, and to a basic [-1, 1] normalization. In addition, we applied few types of normalizations to the GMM-mean supervectors, in order to improve the performance of the SVM classifier. All comparisons of the speaker verification system had been done in terms of DET curve, EER (Equal Error Rate) and Min. DCF. The best results we achieved were on combined supervector normalizations of Universal Background Model (UBM) Standard Deviation (STD) and [-1, 1] normalization. The type of the MFCC normalization has no big influence on the verification performance. The best results were: EER about 5.0% and MIN. DCF of 0.02.

Speaker diarization is ordinarily utilized as a part of the discourse acknowledgment application; it is portrayed in the writing. Then again, utilization of speaker homogeneous bunches, keeping in mind the end goal to adjust the acoustic model to speaker subordinate programmed discourse acknowledgment (ASR) framework that permits enhancing the acknowledgment execution. The i-vectors are likewise called as intermediate vectors. The progression of discourse bunching is as per the following: change over the discourse information in Mel-frequency cepstral coefficient (MFCC) form, model speaker voice by Gaussian mixture model (GMM), adapt the speaker’s expression with normal speaker display, apply the VBEM-GMM, find contrast highlight of a few speakers information, and discover grouping comes about by utilizing traditional k-implies calculation. The Universal Background Model is a huge GMM, and this UBM presents the normal speaker display. The speaker’s information which was at first changed over into crude shape Mel-frequency cepstral coefficient (MFCC) is utilized for speaking to the speaker voice characteristic. The progression of MFCC which is changing is finished by hidden Markov toolkit (HTK).

With the advancement in technology, communication between people around the world from different linguistic backgrounds is increasing gradually, resulting in the requirement of language identification services. Language identification techniques extract distinguishable information as features of a language from the speech corpora to differentiate one language from other. Without publicly available speech corpora, comparison between different techniques will not be much reliable. This paper investigates state-of-the-art features and techniques for language identification of under-resource and closely related languages, namely Pashto, Punjabi, Sindhi, and Urdu. For language identification, speech corpus is designed and collected for mentioned languages. The dataset is a read speech data collected over telephone network (mobile and landline) from different regions of Pakistan. The speech corpus is annotated at the sentence level using X-SAMPA, its orthographic transcription is also provided, and verified data are divided into training and evaluation sets. Mel-frequency cepstral coefficients and their shifted delta cepstral features are used to develop language identification system of target languages. Gaussian mixture model with universal background model (GMM-UBM)-based and I-vector-based language identification approaches are investigated. The results show that GMM-UBM is more effective than the I-vector for language identification of short duration test utterances.

A text-dependent speaker verification system has been implemented for forensic application since 2008 when speech recording is allowed as legal evidence in Indonesian court. Due to the laborious procedures of the current system, an automatic text-independent speaker verification system is being developed in order to accelerate the forensic speaker verification process. The new forensic system is developed using a database spoken in Indonesian. The database is built on two scenarios, interview and conversation. This is done to simulate the real forensic condition for speaker verification in Indonesia. The system adopts the very famous Gaussian Mixture Model (GMM) and Universal Background Model (UBM) framework in speaker verification field and using Mel-Frequency Cepstral Coefficients (MFCC) as the speech feature. Along with the application of zero-normalization, the new forensic speaker verification system has successfully reached an Equal Error Rate (EER) of 4.66 % that is a better performance than the previous developed system of Indonesian forensic speaker verification.

摘要: 在高斯图特征提取过程中,通用背景模型(Universal background model, UBM) 方法常用于根据总体分布估计每一幅图像中特征点分布的高斯混合模型(Gaussian mixture model, GMM)参数. 然而UBM估计的GMM权重参数中有很多接近零的数值,它们所对应的高斯分量对分布估计贡献小却又都参与了计算, 因此UBM的时间复杂度较高. 为解决这个问题,本文提出Bayes UBM方法. 通过引入受限的对称Dirichlet分布来描述GMM权重参数的先验分布,利用Bayes最大后验概率对GMM参数集进行估计. 实验表明Bayes UBM方法不仅有效地降低了时间复杂度,而且提高了Corel数据集上的图像标注精度. 关键词: 图像标注 / 通用背景模型 / 高斯混合模型 / 贝叶斯估计

We propose a new dimension reduction algorithm in combination with the Gaussian Mixture Model (GMM) based Patch Distribution Feature for human gait recognition. Instead of representing each average silhouette image as its gray-level feature, we first extract local patch features at every pixel of the average silhouette image and train a GMM to describe the distribution of the patches in each image. A Universal Background Model (UBM) is first trained with local patch features from all gallery images, then every gallery or probe image is represented by the distribution parameters (referred to as Patch Distribution Features (PDF)) of the image-specific GMM adapted from the UBM. To cope with the high dimension of the PDF feature, the Regularized Trace Ratio Discriminant Analysis (RTRDA) is developed to find the most discriminant subspaces for gait recognition. Experiments on USF humanID database show that RTRDA significantly outperforms the existing algorithms and achieves the best recognition results among all the previous works on USF humanID database in terms of average rank-1 recognition rate.

In this paper, a universal background model (UBM) reconstruction approach is presented, by means of modeling the acoustic space hierarchically and then associating each leaf node of the tree with the most relevant Gaussians. This data-driven algorithm is investigated to potentially utilize broad phonetic-specific speaker characteristics by Gaussian mixture model (GMM). We introduce cumulative posterior probability based Gaussian selection and distance measure based UBM parameter estimation using the tree structure. Analysis and validation of the UBM generated by Gaussian tying are provided in this paper. Performance evaluations are conducted on GMM-supervector and I-vector speaker verification systems, respectively. The experimental results on the NIST SRE 2006 show that the performance can be improved consistently compared to the baseline system.

Evaluation of MSR Identity Toolbox under conditions reflecting those of a real forensic case (forensic_eval_01)

In the article, an algorithm with GMM-UBM (Gaussian mixture model - universal background model) machine learning framework adopted as the cognitive center implemented in a Smart Robot is demonstrated. Moreover, the developed robot is except designed with some embedded smart sensors, the robot is embedded with machine learning technology for applying to the agriculture research field. Specifically, the presented robot could help to analyze the environmental conditions for different plants, e.g. the estimation to weather and humidity, and protection plant from disease destroy. The GMM-UBM algorithm deployed in the smart robot is mainly to control the assignments' behavior precisely. There three of the smart robot are combining with AI (Artificial intelligence) techniques consists of the following equipments: 1) a control movement subsystem, 2) a sensor control subsystem, and 3) an analysis subsystem. The results from the simulation determined the condition of the message options with tag sensing techniques. Moreover, the results have validated that the illustrated system can obtain significantly processing efficiency. Furthermore, the analytic data comes from the analysis subsystem is able to predate the path for the robot move corresponding to the specified conditions.

Automatic speaker identification techniques are widely used nowadays in forensic applications, but its accuracy harshly drops when the voice of the speaker of interest is immersed in a recording containing more than one voice, common situation of investigations where the targets voice are obtained through ambient recordings. In forensic applications where microphones are hidden, such interferent sound sources in recordings are common and they degrade severely the performance of speaker identification techniques. In this paper, we propose a method to mitigate this problem by spatially separating the voice of each speaker using a Blind Source Separation technique called Convolutive Independent Component Analysis, and then applying the separated speech signals to a speaker identification system based on Mel Frequency Cepstral Coefficients and Gaussian Mixture Models. For identifying more than one speaker, the proposed system has a better accuracy than the state-of-the-art solutions.

In this paper, we are introducing speech database consists of 27 Indian languages for analyzing language specific information present in speech. In the context of Indian languages, systematic analysis of various speech features and classification models in view of automatic language identification has not performed, because of the lack of proper speech corpus covering majority of the Indian languages. With this motivation, we have initiated the task of developing multilingual speech corpus in Indian languages. In this paper spectral features are explored for investigating the presence of language specific information. Melfrequency cepstral coefficients (MFCCs) and linear predictive cepstral coefficients (LPCCs) are used for representing the spectral information. Gaussian mixture models (GMMs) are developed to capture the language specific information present in spectral features. The performance of language identification system is analyzed in view of speaker dependent and independent cases. The recognition performance is observed to be 96% and 45% respectively, for speaker dependent and independent environments.

In this study, the classification of the speakers according to age and gender was discussed. Age and gender classes were first examined separately, and then by combining these classes a classification with a total of 7 classes was made. Speech signals represented by Mel-Frequency Cepstral Coefficients (MFCC) and delta parameters were converted into Gaussian Mixture Model (GMM) mean supervectors and classified with a Support Vector Machine (SVM). While the GMM mean supervectors were formed according to the Maximum-a-posteriori (MAP) adaptive GMM-Universal Background Model (UBM) configuration, the number of components was changed from 16 to 512, and the optimum number of components was decided. Gender classification accuracy of the system developed using aGender dataset was measured as 99.02% for two classes and 92.58% for three classes and age group classification accuracy was measured as 67.03% for female and 63.79% for male. In the classification of age and gender classes together in one step, an accuracy of 61.46% was obtained. In the study, a two-level approach was proposed for classifying age and gender classes together. According to this approach, the speakers were first divided into three classes as child, male and female, then males and females were classified according to their age groups and thus a 7-class classification was realized. This two-level approach was increased the accuracy of the classification in all other cases except when 32-component GMMs were used. While the highest improvement of 2.45% was achieved with 64 component GMMs, an improvement of 0.79 was achieved with 256 component GMMs.

Spoken language identification is a field of research that is already being done by many people. There are many techniques proposed for doing speech processing, such as Support Vector Machines, Gaussian Mixture Models, Decision Trees, and others. This paper will use the system using the Mel-Frequency Cepstral Coefficient (MFCC) features of speech input signal, use Random Forest (RF), Gaussian Mixture Model (GMM), and K-Nearest Neighbor (KNN) as a classifier, use the 3s, 10s, and 30s as scoring method, and use dataset that consists of Javanese, Sundanese, and Minang languages which are traditional languages from Indonesia. K-Nearest Neighbor has 98.88% of accuracy for 30s of speech and followed by Random Forest that has 95.55% of accuracy for 30s of speech, GMM has 82.24% of accuracy.

This work proposes a novel method of predicting formant frequencies from a stream of mel-frequency cepstral coefficients (MFCC) feature vectors. Prediction is based on modelling the joint density of MFCCs and formant frequencies using a Gaussian mixture model (GMM). Using this GMM and an input MFCC vector, two maximum a posteriori (MAP) prediction methods are developed. The first method predicts formants from the closest, in some sense, cluster to the input MFCC vector, while the second method takes a weighted contribution of formants predicted from all clusters. Experimental results are presented using the ETSI Aurora connected digit database and show that predicted formant frequencies are within 3.2% of reference formant frequencies.