Mel-frequency Discrete Wavelet Coefficients Research Articles

Speech Emotion Classification and Recognition with different methods for Turkish Language

In several
 application, emotion  recognition from
 the speech signal has been research topic since many years. To determine the
 emotions from the speech signal, many systems have been developed. To solve the
 speaker emotion recognition problem, hybrid model is proposed to classify five
 speech emotions, including  anger,
 sadness, fear, happiness and neutral. The aim this study of was to actualize
 automatic voice and speech emotion recognition system using hybrid model taking
 Turkish sound forms and properties into consideration.  Approximately 3000 Turkish voice samples of
 words and clauses with differing lengths have been collected from 25 males
 and  25 females. In this study, an
 authentic and unique  Turkish  database has been used. Features of these
 voice samples have been obtained using Mel Frequency Cepstral Coefficients
 (MFCC) and Mel Frequency Discrete Wavelet Coefficients (MFDWC). Moreover,
 spectral features of these voice samples have been obtained  using Support Vector Machine (SVM). Feature
 vectors of the voice samples obtained have been trained with such methods as
 Gauss Mixture Model( GMM), Artifical Neural Network (ANN), Dynamic Time Warping
 (DTW), Hidden Markov Model (HMM) and hybrid model(GMM with combined SVM).  This hybrid model has been carried out by
 combining with SVM and GMM.  In first
 stage of this model, with SVM has been performed  subsets obtained vector of  spectral features. In the second  phase, a set of training and tests have been
 formed from these spectral features. In the test phase, owner of a given voice
 sample has been identified taking the trained voice samples into consideration.
 Results and performances of the algorithms employed in the study for
 classification have been also demonstrated in a comparative manner.         

Speech recognition system for Turkish language with hybrid method

Currently, technological developments are accompanied by a number of associated problems. Security takes the first place amongst such problems. In particular, biometric systems such as authentication constitute a significant fraction of the security problem. This is because sound recordings having connection with various crimes are required to be analysed for forensic purposes. Authentication systems necessitate transmission, design and classification of biometric data in a secure manner. The aim of this study is to actualise an automatic voice and speech recognition system using wavelet transform, taking Turkish sound forms and properties into consideration. Approximately 3740 Turkish voice samples of words and clauses of differing lengths were collected from 25 males and 25 females. The features of these voice samples were obtained using Mel-frequency cepstral coefficients (MFCCs), Mel-frequency discrete wavelet coefficients (MFDWCs) and linear prediction cepstral coefficient (LPCC). Feature vectors of the voice samples obtained were trained with k-means, artificial neural network (ANN) and hybrid model. The hybrid model was formed by combining with k-means clustering and ANN. In the first phase of this model, k-means performed subsets obtained with voice feature vectors. In the second phase, a set of training and tests were formed from these sub-clusters. Thus, for being trained more suitable data by clustering increased the accuracy. In the test phase, the owner of a given voice sample was identified by taking the trained voice samples into consideration. The results and performance of the algorithms used for classification are also demonstrated in a comparative manner.
 Keywords: Speech recognition, hybrid model, k-means, artificial neural network (ANN).

Automatic Voice and Speech Recognition System for the German Language with Deep Learning Methods

In our age, technological developments are accompanied by certain problems associated with them. Security takes the first place amongst such kind of problems. In particular, such biometric systems as authentication constitute the significant fraction of the security matters. This is because sound recordings having connection with the various crimes are required to be analyzed for forensic purposes. Authentication systems necessitate transmission, design and classification of biometric data in a secure manner. In this study, analysis of German language employed in the economy, industry and trade in a wide spread manner, has been performed. In the same vein, the aim was to actualize automatic voice and speech recognition system using Mel Frequency Cepstral Coefficients (MFCC), MelFrequency Discrete Wavelet Coefficients (MFDWC) and Linear. Prediction Cepstral Coefficient (LPCC) taking German sound forms and properties into consideration. Approximately 2658 German voice samples of words and clauses with differing lengths have been collected from 50 males and 50 females. Features of these voice samples have been obtained using wavelet transform. Feature vectors of the voice samples obtained have been trained with such methods as Boltzmann Machines and Deep Belief Networks. In the test phase, owner of a given voice sample has been identified taking the trained voice samples into consideration. Results and performances of the algorithms employed in the study for classification have been also demonstrated in a comparative manner.

Experimentation Using Short-Term Spectral Features for Secure Mobile Internet Voting Authentication

We propose a secure mobile Internet voting architecture based on the Sensus reference architecture and report the experiments carried out using short-term spectral features for realizing the voice biometric based authentication module of the architecture being proposed. The short-term spectral features investigated are Mel-Frequency Cepstral Coefficients (MFCCs), Mel-Frequency Discrete Wavelet Coefficients (MFDWC), Linear Predictive Cepstral Coefficients (LPCC), and Spectral Histogram of Oriented Gradients (SHOGs). The MFCC, MFDWC, and LPCC usually have higher dimensions that oftentimes lead to high computational complexity of the pattern matching algorithms in automatic speaker recognition systems. In this study, higher dimensions of each of the short-term features were reduced to an 81-element feature vector per Speaker using Histogram of Oriented Gradients (HOG) algorithm while neural network ensemble was utilized as the pattern matching algorithm. Out of the four short-term spectral features investigated, the LPCC-HOG gave the best statistical results withRstatistic of 0.9127 and mean square error of 0.0407. These compact LPCC-HOG features are highly promising for implementing the authentication module of the secure mobile Internet voting architecture we are proposing in this paper.

Applied mel-frequency discrete wavelet coefficients and parallel model compensation for noise-robust speech recognition

Interfering noise severely degrades the performance of a speech recognition system. The Parallel Model Compensation (PMC) technique is one of the most efficient techniques for dealing with such noise. Another approach is to use features local in the frequency domain, such as Mel-Frequency Discrete Wavelet Coefficients (MFDWCs). In this paper, we investigate the use of PMC and MFDWC features to take advantage of both noise compensation and local features (MFDWCs) to decrease the effect of noise on recognition performance. We also introduce a practical weighting technique based on the noise level of each coefficient. We evaluate the performance of several wavelet-schemes using the NOISEX-92 database for various noise types and noise levels. Finally, we compare the performance of these versus Mel-Frequency Cepstral Coefficients (MFCCs), both using PMC. Experimental results show significant performance improvements for MFDWCs versus MFCCs, particularly after compensating the HMMs using the PMC technique. The best feature vector among the six MFDWCs we tried gave 13.72 and 5.29 points performance improvement, on the average, over MFCCs for −6 and 0 dB SNR, respectively. This corresponds to 39.9% and 62.8% error reductions, respectively. Weighting the partial score of each coefficient based on the noise level further improves the performance. The average error rates for the best MFDWCs dropped from 19.57% to 16.71% and from 3.14% to 2.14% for −6 dB and 0 dB noise levels, respectively, using the weighting scheme. These improvements correspond to 14.6% and 31.8% error reductions for −6 dB and 0 dB noise levels, respectively.