PRAYAS - an International Journal of Multidisciplinary Studies
Total Page:16
File Type:pdf, Size:1020Kb
PRAYAS - An International Journal of Multidisciplinary Studies Dear Readers, I take immense pleasure to present the Prayas - An international journal of multidisciplinary studies, PIJMS II (1). This peer-reviewed journal is the first attempt of its kind in the state of Tripura. I hope this journal will enthuse high quality research activities in the state and will encourage the readers as well. I feel mere gathering knowledge is not sufficient unless and until there is ample endeavour to apply the gathered knowledge. Finally I congratulate the whole team led by Mr. Basu Maan Daas that made the journal PIJMS a grand success. I hope you all will appreciate the achievement and wish the journal a great future. Dr. Sambhunath Rakshit Principal, Government Degree College, Dharmanagar, Tripura, India. PIJMS II (1) – 2015 ONLINE ISSN No. : 2348-618X PRAYAS - An International Journal of Multidisciplinary Studies Dear readers, Prayas - An international journal of multidisciplinary studies (PIJMS) is the first open access, international, peer-reviewed, annual research journal of multidisciplinary studies from Tripura intended for Professors, academicians, scholars, professionals and students engaged in research. The main purpose of the journal will be to encourage research publication and provide an international forum to disseminate knowledge. The initiative is taken by the Government Degree College, Dharmanagar, Tripura. As the Editor-in-chief of the journal, I accepted the challenge and put in my utmost efforts to make the journal possible. I thank my Principal, Dr. S. Rakshit and other colleagues, who had shown immense faith in me. I acknowledge other members of the Literary Committee, who gave me full freedom to shape the journal as I could. Finally, I would like to thank the all eminent advisers and referees, who responded to my request, encouraged me and bore the strain to pick the right manuscripts and help the authors to edit them, whenever necessary. Without the support of each and every one associated the journal would have been impossible. I congratulate the authors whose papers are eventually selected by the blind referees and I am deeply sorry to those whose papers we couldn’t include. I hope in near future PIJMS would attain high impact factor and good number of citations. Kindly wish the journal success. Let the spirit never die … Let it consistently dive and fly … With no sign whatever of any sigh … In the overhanging, looming sky … With gloomy clouds passing by ... Basu Maan Daas Editor-in-chief PIJMS Assistant Professor, Department of Chemistry Government Degree College, Dharmanagar, Tripura, India. PIJMS II (1) – 2015 ONLINE ISSN No. : 2348-618X PRAYAS - An International Journal of Multidisciplinary Studies A COMPARATIVE STUDY OF DIFFERENT FEATURE VECTOR FOR RECOGNITION OF ISOLATED ASSAMESE AND BODO WORDS C. Saloi, Page 1 A STUDY ON THE TREND OF ENROLLMENT AND PERFORMANCE OF THE FEMALE STUDENTS IN SCIENCE AND TECHNICAL COURSES OFFERED BY DIFFERENT TECHNICAL COLLEGES OF AGARTALA B. Saha, H. Mukherjee & M. M. D. Biswas, Page 16 ARBUSCULAR MYCORRHIZAL FUNGI IN TROPICAL FOREST ECOSYSTEMS – POTENTIALS AND APPLICATIONS K. N. Singh, S. Ahanthem & D. K. Jha, Page 22 COMPARISON OF EXPLOSIVE LEG STRENGTH BETWEEN FEMALE TRIBAL AND NON - TRIBAL STUDENTS OF TRIPURA A. Sinha , Page 34 INTRODUCING ICT IN EDUCATION - HIGH SCHOOL TEACHERS’ PERSPECTIVES TO THE CHANGING CLASSROOM PRACTICES S. Paul & S. K. Rath, Page 37 MICRO-INSURANCE - A STEP TOWARDS INCLUSION OF ECONOMICALLY POOR TO THE UMBRELLA OF INSURANCE P. Shil & B. D. Nath, Page 49 NARRATIVES OF GABRIEL GARCÍAMÁRQUEZ BETWEEN CENSORSHIP AND SCANDAL M. Boro, Page 58 OBESITY – THE ESCALATING GLOBAL EPIDEMIC - A REVIEW S. Chatterjee, Page 67 PARTICIPATION OF MUSLIM WOMEN IN HIGHER EDUCATION S. Ahmed & N. Sultana, Page 78 PIJMS II (1) – 2015 ONLINE ISSN No. : 2348-618X PRAYAS - An International Journal of Multidisciplinary Studies VIOLENCE AGAINST WOMAN IS A BURNING ISSUE - A SOCIAL CRIME S. Pal & M. Paul, Page 85 WOMEN SECURITY IN MODERNISED SOCIETY B. Paul, Page 99 PIJMS II (1) – 2015 ONLINE ISSN No. : 2348-618X PRAYAS - An International Journal of Multidisciplinary Studies A COMPARATIVE STUDY OF DIFFERENT FEATURE VECTOR FOR RECOGNITION OF ISOLATED ASSAMESE AND BODO WORDS Chandan Saloi, Department of Instrumentation & USIC, Gauhati University, Guwahati. Abstract Feature extraction process is a technique of acquiring a priori knowledge used to transform an input in the signal space to an output in a feature space to achieve some desired criteria. If lots of clusters in a high dimensional space must be classified, the objective of feature extractor is to transform that space such that classifying becomes easier. The feature extractor block designed in speech recognition aims towards reducing the complexity of the problem before the next stage start to work with the data. This chapter describes how to extract information from a speech signal, representing in the form of feature vectors. LPC Cepstral coefficients are used to detect the basic phonemes of a language. To recognise any words of Assamese & Bodo language, it is very important to detect the various features of both Assamese & Bodo languages. But before recognition, the correct cepstral information is very important. But we have seen that while we are recognizing the phoneme of the Bodo language, it is observed that only LPC cepstral coefficient is not sufficient to detect the phonemes that have similar spectral features but distinct energy contents and pitch period. This is a common phenomenon in tonal languages. Bodo is a tonal language and therefore, to recognize the tone, frame energy and first order derivatives of frame energy along with pitch period and its first and second order derivatives are used. Keywords: LPC cepstral coefficients, pre-emphasis ******* Feature extraction The purpose of feature extraction in speech recognition is to transform speech signals into a set of vectors essential for speech recognition while discarding unreliable parts. The transformed vector is called a feature vector. Several representations of feature vectors have been used for speech recognition, which usually belong to one of the following four categories [1]: linear predictive coding (LPC)-related parameters, parameters from the filter bank analysis, phonetically important parameters such as formants, energy, zero crossing rates, voice on-set time, and parameters based on the auditory modelling. In the present work, LPC-based cepstral coefficients and phonetically important parameters are used as feature vectors. The schematic diagram of feature extraction method is shown in Figure 1. Fig 1: Feature Extractor schematic diagram PIJMS II (1) – 2015 ONLINE ISSN No. : 2348-618X Page 1 PRAYAS - An International Journal of Multidisciplinary Studies Speech sampling A speech signal is first low-pass filtered to prevent the aliasing effect in sampling and removing parts of speech spectrum which are not important and may contain noise. It might be argued that a higher sampling frequency, or more sampling precision, is needed to achieve higher recognition accuracy. However, if a normal digital phone, which samples speech at 8,000 Hertz with an 8 bit precision, is able to preserve most of the information carried by the signal [2], it does not seem necessary to increase the sampling rate beyond 11,025 Hertz or the sampling precision to something higher than 8 bits. Another reason behind these settings is that commercial speech recognizers typically use comparable parameter values and achieve impressive results. In this work, the incoming signal is sampled at 8 kHz with 16 bits of precision shown in Figure 2. This is because the speech to be recognized by the proposed system includes not only voiced speech but also unvoiced such as /s/. Therefore, higher sampling frequency and more sampling precision is a must. The speech is recorded and sampled using an off-the-shelf relatively inexpensive dynamic microphone and a standard PC sound card by using the software CoolEdit 2000. Fig 2: Speech signal for the word ‘aalu’ (top-Assamese, bottom-Bodo) sampled at 8 kHz with a precision of 16 bits. Pre-emphasis In general, the digitized speech waveform has a high dynamic range and suffers from additive noise. For this reason, pre-emphasis is applied to spectrally flatten the signal in order to make it less susceptible to finite precision effects (such as overflow and underflow) later in the speech processing. The most widely used pre-emphasis is the fixed first- order system. The calculation of pre-emphasis is shown in equation (1). H (z) = 1− az-1 0.9 ≤ a ≤ 1.0 (1) The most common value for a, is 0.95 (Deller et al., 1993) [3]. A pre-emphasis can be expressed as given by equation (2). Š(n) = s(n) – 0.95 s(n-1) (2) Fig 3: Pre-emphasised signal in Time & Frequency domain for the Assamese word ‘aalu’ PIJMS II (1) – 2015 ONLINE ISSN No. : 2348-618X Page 2 PRAYAS - An International Journal of Multidisciplinary Studies Fig 4: Pre-emphasised signal in Time & Frequency domain for the Assamese word ‘maina’ Fig 5: Pre-emphasised signal in Time & Frequency domain for the Bodo word ‘aalu’ Fig 6: Pre-emphasised signal in Time & Frequency domain for the Bodo word ‘maina’ Frame blocking The speech signal is dynamic or time-variant in the nature. According to Rabiner (1993), the speech signal is assumed to be stationary when it is examined over a short period of time. In order to analyse the speech signal, it has to be blocked into frames of N samples, with adjacent frames being separated by M samples as shown in Figure 7. In other words, the frame is shifted with M samples from the adjacent frame. If the shifting is small, then the LPC spectral estimated from frame to frame will be very smooth. If there is no overlapping between adjacent frames, the speech signals will be totally lost and correlation between the result LPC spectral estimates of adjacent frames will contain a noisy component.