WAVELET DETAIL COEFFICIENT AS A NOVEL WAVELET-MFCC FEATURES IN TEXT-DEPENDENT SPEAKER RECOGNITION SYSTEM

Speaker recognition is the process of recognizing a speaker from his speech. This can be used in many aspects of life, such as taking access remotely to a personal device, securing access to voice control, and doing a forensic investigation. In speaker recognition, extracting features from the speech is the most critical process. The features are used to represent the speech as unique features to distinguish speech samples from one another. In this research, we proposed the use of a combination of Wavelet and Mel Frequency Cepstral Coefficient (MFCC), Wavelet-MFCC, as feature extraction methods, and Hidden Markov Model (HMM) as classification. The speech signal is first extracted using Wavelet into one level of decomposition, then only the sub-band detail coefficient is used as the feature for further extraction using MFCC. The modeled system was applied in 300 speech datasets of 30 speakers uttering “HADIR” in the Indonesian language. K-fold cross-validation is implemented with five folds. As much as 80% of the data were trained for each fold, while the rest was used as testing data. Based on the testing, the system's accuracy using the combination of Wavelet-MFCC obtained is 96.67%. ABSTRAK: Pengecaman penutur adalah proses mengenali penutur dari ucapannya yang dapat digunakan dalam banyak aspek kehidupan, seperti mengambil akses dari jauh ke peranti peribadi, mendapat kawalan ke atas akses suara, dan melakukan penyelidikan forensik. Ciri-ciri khas dari ucapan merupakan proses paling kritikal dalam pengecaman penutur. Ciri-ciri ini digunakan bagi mengenali ciri unik yang terdapat pada sesebuah ucapan dalam membezakan satu sama lain. Penyelidikan ini mencadangkan penggunaan kombinasi Wavelet dan Mel Frekuensi Pekali Cepstral (MFCC), Wavelet-MFCC, sebagai kaedah ekstrak ciri-ciri penutur, dan Model Markov Tersembunyi (HMM) sebagai pengelasan. Isyarat penuturan pada awalnya diekstrak menggunakan Wavelet menjadi satu tahap penguraian, kemudian hanya pekali perincian sub-jalur digunakan bagi pengekstrakan ciri-ciri berikutnya menggunakan MFCC. Model ini diterapkan kepada 300 kumpulan data ucapan daripada 30 penutur yang mengucapkan kata "HADIR" dalam bahasa Indonesia. Pengesahan silang K-lipat dilaksanakan dengan 5 lipatan. Sebanyak 80% data telah dilatih bagi setiap lipatan, sementara selebihnya digunakan sebagai data ujian. Berdasarkan ujian ini, ketepatan sistem yang menggunakan kombinasi Wavelet-MFCC memperolehi 96.67%.

Implementation of MODWT on Real Time Power Quality Disturbance Signal

Maximal overlap discrete wavelet transform (MODWT) is the upgradation of traditional wavelet transform (WT), has been employed for localization of different power quality disturbance signal (PQDS). Every signal has been break down up to fourth level to localize the disturbances. The co-efficient of MODWT have been again employed for classification. The selected indices have been obtained utilizing the detail coefficient of this variant of WT. These features are the inputs to the data mining classifier. Decision Tree (DT) have been implemented for discrimination of PQ disturbance signals. Various PQDS have been generated in noisy and noise free climate. Besides this, the aforementioned techniques is examined with three phase signals bring out from transmission line panels.

Feature Extraction of Facial Electromyograph (EMG) Signal for Aceh Languages Speech using Discrete Wavelet Transform (DWT)

The facial electromyograph (FEMG) signal is a signal that occurs in the muscles of the contracted human face. This FEMG signal is one of the techniques used to study human speech recognition. It can be acquired by placing an electrode surface on the skin around the facial articulation muscle. Three types of muscles in this study are the masseter, risorius and depressor muscle. This study aims to extract and analyze the features in the FEMG signal. The extraction method is the discrete wavelet transform (DWT). The type of wavelet transform is Daubechies2 with level 5. After extraction and analysis of FEMG signals, the FEMG signal pattern for each spoken word indicated by differences in the approximation and detail coefficient of the FEMG signal. In addition, the level of difference in the FEMG signal pattern is also indicated by the histogram of the approximation coefficient of the FEMG signal. Thus, the discrete wavelet transform method can be used as one of the methods for extracting the FEMG signal feature in a human facial electromyograph (FEMG) signal.

Multifocus image fusion based on coefficient significance of redundant discrete wavelet transform

Purpose This study aims to obtain the more precise decision map to fuse the source images by Coefficient significance method. In the area of multifocus image fusion, the better decision map is very important the fusion results. In the processing of distinguishing the well-focus part with blur part in an image, the edge between the parts is more difficult to be processed. Coefficient significance is very effective in generating the better decision map to fuse the multifocus images. Design/methodology/approach The energy of Laplacian is used in the approximation coefficients of redundant discrete wavelet transform. On the other side, the coefficient significance based on statistic property of covariance is proposed to merge the detail coefficient. Findings Due to the shift-variance of the redundant discrete wavelet and the effectiveness of fusion rule, the presented fusion method is superior to the region energy in harmonic cosine wavelet domain, pixel significance with the cross bilateral filter and multiscale geometry analysis method of Ripplet transform. Originality/value In redundant discrete wavelet domain, the coefficient significance based on statistic property of covariance is proposed to merge the detail coefficient of source images.

Performance Evaluation of Traveling Wave Fault Locator for a 220kV Hoa Khanh-Thanh My Transmission Line

This paper presents the traveling wave based fault location methods of SEL-400L, and SFL-2000 available on the market for a 66.9km, 220kV Hoa Khanh-Thanh My transmission line in Central Viet Nam, such as single-ended, and double-ended, all of which rely on measurements from inductive CTs and capacitive VTs. Focus was given on the building process of a Matlab Simulink model to evaluate these methods. Current and voltage signals were sent to an analog Chebyshev type II filter which passes higher frequency signals at 3kHz and rejects low frequencies signal at 50Hz. After that, these output signals are used in Clarke's transformation for getting 0 and α components. The detail coefficient of the selected components after DWT using Db4 wavelet at decomposition level 1 can be used to determine the fault types, the direction of fault and propose a crest-wave comparison solution to identify exactly the adjacent bus' reflected wave from the fault point's reflected wave for the fault location. Finally, the accuracy of fault location on the transmission line is reviewed by varying various parameters like fault type, fault location and fault resistance on a given power system model.

WAVELET-BASED TRANSMISSION LINE FAULT DETECTION AND CLASSIFICATION

An appropriate method of fault detection and classification of power system transmission line using discrete wavelet transform is proposed in this paper. The detection is carried out by the analysis of the detail coefficients energy of phase currents. Discrete Wavelet Transform (DWT) analysis of the transient disturbance caused as a result of occurrence faults is performed. The work represented in this paper is focused on classification of simple power system faults using the maximum detail coefficient, energy of the signal and the ratio of energy change of each type of simple simulated fault are characteristic in nature and used for distinguishing fault types.

Image Denoising Using Wavelet Transform and Wavelet Transform with Enhanced Diversity

This paper is a comparative study of image denoising using previously known wavelet transform and new type of wavelet transform, namely, Diversity enhanced discrete wavelet transform. The Discrete Wavelet Transform (DWT) has two parameters: the mother wavelet and the number of iterations. For every noisy image, there is a best pair of parameters for which we get maximum output Peak Signal to Noise Ratio, PSNR. As the denoising algorithms are sensitive to the parameters of the wavelet transform used, in this paper comparison of DEDWT to DWT has been presented. The diversity is enhanced by computing wavelet transforms with different parameters. After the filtering of each detail coefficient, the corresponding wavelet transforms are inverted and the estimated image, having a higher PSNR, is extracted. To benchmark against the best possible denoising method three thresholding techniques have been compared. In this paper we have presented a more practical, implementation oriented work.

The Method of Logging Curves Fusion Based on Multi-Scale Wavelet Transform

The mutation and form of logging curve can be represented by the variation of modulus maxima coefficients of wavelet transform within different scales exactly, then we can use wavelet multiscale edge detection theory to analyze characteristics of sequence stratigraphy boundaries in logging curves. Obtain the modulus maximum of approximation coefficient matrix and detail coefficient matrix after decompositing GR and SP curve in every scales.Compare and amend the modulus maximum of approximation coefficient matrix and detail coefficient matrix reciprocally, applicate the Mallat alternation foldover algorithm to reconstruction logging curve eventually ,we can get the fusion curves in different scales. The fusion curves can greatly enhance characteristics of sequence stratigraphy boundaries in logging curves.

Application of Wavelet-Based Fractal Dimension Threshold Denoising Method to Load Time History of Engineering Vehicle

The load time history signal of engineering vehicle is usually polluted by various nonstationary and stationary noises in the field test. An approach based on wavelet transform (WT) and fractal dimension (FD) is proposed in order to improve the adaptability and efficiency of denoising. This method initially decomposes the original signal into detail and approximation space in the WT domain by WT-based multiresolution decomposition. The short-time fractal dimension of detail coefficient is calculated at each scale. After the application of the binary processing to the short-time fractal dimensions, the locations where the thresholding of the detail coefficients has to be executed are ensured. The desired load signal is provided by applying WT-based multiresolution reconstruction to the processed detail coefficients and the unprocessed approximation coefficients. The proposed method is applied to an actual load time history signal of engineering vehicle. And the performance of this method is compared with that of the WT-based hard thresholding denoising method. The results show that this method is an alternative way to process the load time history signal of engineering vehicle.