UTPAL
@rgu.ac.in
Professor, Department of Computer Science and Engineering
Rajiv Gandhi University
RESEARCH, TEACHING, or OTHER INTERESTS
Computer Science
Scopus Publications
Scopus Publications
Bomken Kamdak, Gom Taye, and Utpal Bhattacharjee
Springer Nature Singapore
Hem Chandra Das and Utpal Bhattacharjee
IEEE
The novel approach of this paper is the introduction of a novel information blending paradigm. Two basic classifiers are first trained using the shifting delta cepstra (SDC) and prosodic features, respectively. To improve Assamese dialect detection accuracy, the co-training approach is used in semi-supervised learning. This system evaluates four Assamese dialects. The experimental data show that the suggested system performs better than the existing system that used GMM.
Hem Chandra Das and Utpal Bhattacharjee
Springer Nature Singapore
Jyoti Mannala, Bomken Kamdak, and Utpal Bhattacharjee
Springer Nature Singapore
Debashis Dev Misra, Kandarpa Kumar Sarma, Utpal Bhattacharjee, Pradyut Kumar Goswami, and Nikos Mastorakis
IEEE
Ultra Dense Network (UDN) is a guiding principle in the direction of 5G network challenges which focuses on network infrastructure densification. The extremely dynamic nature of such networks implies that the optimal path between the source/destination pairs will be highly dynamic in nature. In such a backdrop, routing mechanisms have to be designed to be robust and scalable in nature in order to ensure seamless link reliability and quality of service of the network. The shortest optimal route of the source/destination pair is found using a combination of evolutionary optimization algorithms namely Genetic Algorithm (GA), Particle Swarm Optimization (PSO) Algorithm and our proposed hybrid PSO-GA which searches for an optimal option by determining cost functions of individual fitness state and comparing states generated between individual solutions. Applying all the three proposed algorithms to the Shortest Path routing problem in UDNs has led us to believe that the hybrid PSO-GA has given a comparatively better solution.
Renu Singh, Utpal Bhattacharjee, Arvind Kumar Singh, and Madhusudhan Mishra
Springer Singapore
Renu Singh, Arvind Kumar Singh, and Utpal Bhattacharjee
Springer Singapore
Renu Singh, Utpal Bhattacharjee, and Arvind Kumar Singh
Elsevier BV
A Firos and Utpal Bhattacharjee
IEEE
Speech encoding refers to compression for transmission or storage, possibly to an unintelligible state, with decompression used prior to playback. This paper attempts to formulate the semantic level compression technique on speech signals by preserving its prosodic features. LPC analysis will be done to identify the feature of the input speech. GMM will be used to preserve the emotional content while encoding. ANN will be utilized to identify the best features for encoding. Using such semantic based coding will highly reduce the computational overhead in speech coders.
Ani Taggu, Bikram Patir, and Utpal Bhattacharjee
IEEE
Vehicles of today are increasingly being networked via various available networking technologies. IEEE 802.11p advocates Vehicle-to-Vehicle and Vehicle-to-Infrastructure communication via Wireless Access in Vehicular Environments (WAVE) between vehicles in the frequency range of 5.9 GHz. Also, IEEE 802.11j proposes the usage of 4.9 GHz frequency range for Wi-Fi. This paper proposes a dual band antenna that is capable of operating in both the WAVE and Wi-Fi bands. This proposed antenna is expected to be simple, easy-to-produce and inexpensive; it can be a cost-effective alternative to use of multiple directional antennas for vehicles. The choice of microstrip patch antenna technology with defected ground structure (DGS) was driven by cost considerations and ease of bulk manufacturing. This omni-directional antenna is expected to be fitted in a central location in the vehicle to avoid requirement of two or more directional antennas. The proposed antenna is characterized by popular antenna design software Ansoft HFSS.
Swapnanil Gogoi and Utpal Bhattacharjee
IEEE
In this paper, vocal tract length normalization (VTLN) and sub-band spectral subtraction (SSS) have been used for speaker adaptation and noise reduction to develop an Assamese vowel recognition system which is robust to the speaker and environment variabilities. In the present work VTLN has been implemented to reduce the effects of inter speaker variabilities and sub-band spectral subtraction has been used to reduce the effects of environmental variabilities. The effectiveness of VTLN in noisy and noise-free environment has been evaluated for Assamese vowel recognition system. The Assamese vowel recognition system has been implemented using Hidden Markov Model (HMM). Mel Frequency Cepstral Coefficient (MFCC) has been used as feature vector. Experimented result shows that the performance of the system improved considerably after applying VTLN technique in noise-free and some of the noisy conditions.
Utpal Bhattacharjee, Swapnanil Gogoi, and Rubi Sharma
IEEE
Noise is omnipresent in almost all acoustical environments. The investigation presents here seeks to quantify the impact of noise on mel-frequency cepstral coefficients (MFCC) of speech signal. MFCC is one of the most commonly used features for speech recognition systems. However, it has been observed that performance of MFCC based system degrades drastically with changing noise levels and noise types. In the present study, different noise types at different levels have been added to the clean speech signal and the changes in statistical distribution pattern of the signal has been investigated. Further, performance of two commonly used noise normalization techniques Cepstral Mean and Variance Normalization (CMVN) and Spectral Subtraction (SS) have also been evaluated.
Pranab Das and Utpal Bhatacharjee
IEEE
This paper proposes a new approach of noise reduction based on the analysis of MFCC feature space using self-organizing map network. Here the U-matrix plot of the feature space is analyzed in presence of white noise at different signal to noise ratio. Based on the observation, boundary neurons separating clusters are identified in the feature space. For each such neuron in the boundary, its 2-D feature vector is extracted from the U-matrix and hit matrix. This collection of feature vectors based on the boundary neurons are eliminated from the original feature space. Thus the new feature space obtained is used to perform the tasks of visualization and speaker verification. Experiments were carried out by combining synthetic white noise with real world data sets.
Pranab Das and Utpal Bhattacharjee
IEEE
In real world situation performance of speaker verification system drops significantly because of mismatched training and test conditions. In this paper we have analyzed three factors namely noise, channel variability and session variability, that are responsible for poor performance of a speaker verification system. The first step towards noise robustness GFCC features were used as recent research has shown better noise robustness of gammatone frequency cepstral coefficients over mel-frequency cepstral coefficients. In the second step robustness towards session and channel variability is achieved by shifting from the classical way of modeling a speaker to a rather new approach of joint factor analysis. Experimental results over different acoustic environment and over different SNR have shown significant improvement in the performance of the system.
U. Bhattacharjee and K. Sarmah
IEEE
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.
Utpal Bhattacharjee
Springer Berlin Heidelberg
Utpal Bhattacharjee and Kshirod Sarmah
IEEE
This paper report the experiments carried out on the recently collected speaker recognition database to study the impact of language variability on speaker verification system. The speech database consists of speech data recorded from 100 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 of Arunachal Pradesh. The collected database is evaluated with Gaussian mixture model based speaker verification system. The impact of the mismatch in training and testing language has been evaluated. The initial study explores the impact of language mismatched in the training and testing on the performance of the speaker verification system.
Renu Singh, Swanirbhar Majumder, U. Bhattacharjee, and A. Dinamani Singh
Springer Berlin Heidelberg