A New Speaker Recognition System with Combined Feature Extraction Techniques

Abstract

Problem statement:This study introduces a new method for speaker verification system by fusing two different feature extraction methods to improve the recognition accuracy and security. Approach: The proposed system uses Mel frequency cepstral coefficients for speaker identification and Modified MFCC for verification. For speaker modeling vector quantization is used. Results: The proposed system was investigated the effect of the different length segmental feature as well as speaker modeling for speaker recognition. The performance was evaluated against 1000 speakers for 10 different languages with duration of 10 sec for training the system and for testing 5 sec. duration samples were used. Conclusion/Recommendations: Experimental results of the proposed system showed that higher recognition accuracy of 93% is achieved by increasing the number of filter banks used for feature extraction method, more competitive with existing system using vector quantization with lesser computational complexity. The system efficiency may further be improved using other speaker modeling techniques like GMM, HMM.

Machine Learning Based Listener Classification and Authentication Using Frequency Following Responses to English Vowels for Biometric Applications

“…Also, only 2 cases were misclassified as a negative and 6 cases were misclassified to be positive, yielding a precision and recall of 0.89 and 0.73 for the positive class and 0.77 and 0.91 for the negative class, respectively. [12,13,14,15,16,17,18,19,20,21,22] in the two scenarios of Test/Retest and Retest/Test. The results are shown for the RBF, linear and polynomial kernels, respectively.…”

“…This is also in agreement with observations in Chapter 4. Classification results for case 2 in which the sheep class composition included subjects [1,3,5,7,9,11,13,15,17,19,21] and wolves class composition included subjects [2,4,6,8,10,12,14,16,18,20,22] are shown in Tables 5.5, 5.6 and 5.7. The best accuracy for this case was obtained using the RBF kernel (90.91%) which was for the Retest/Test scenario of the highest computational resolution spectrogram (window size of 512 samples and overlap size of 511 samples).…”

“…Speaker recognition has also been shown to be susceptible to failure due to the possibility of creating a mathematical model of a person's speech production system and mimicking their voice. Smith [22] reported that speaker recognition systems can be easily fooled by speech synthesized from a mathematical model of a target individual's speech production system.…”

Frontal Arabic Fricative Consonants Characteristics among Primary School Children: Spectral Density Function Approach

“…The common method to distinguish the speech samples are spectrogram reading which proved to be the powerful method to obtain the formants of each samples and had been used in long time ago in speech processing until now (Zue and Cole, 1979;Johannsen et al, 1983;Hassan, 1982;Zue and Lamel, 1986;Debyeche et al, 1998;Johnson, 2003;Awadalla et al, 2005;Awais et al, 2006;Steinberg and O'Shaughnessy, 2008;Iqbal et al, 2008;Abdul-Kadir et al, 2010, Mporas et al, 2007, Sumithra et al, 2011, Abdul-Kadir et al, 2011a. Dental, (3) Alveolar, (4) Post-alveolar, (5) Palatal, (6) Velar, (7) Pharyngeal, (8) Glottis…”

An Effective System for Acute Spotting Aberration in the Speech of Abnormal Children Via Artificial Neural Network and Genetic Algorithm

“…Normal speech waveform may vary from time to time depending on the physical condition of speakers' vocal cord. Rather than the speech waveforms themselves, MFFCs are less susceptible to the said variations (Sumithra et al, 2011). The following steps are involved in extracting the MFCC feature (Bharathi and Shanthi, 2011): The aforesaid eqns are utilized to obtain the PCA of both M a and M b here given eqns are the general sets of eqns to generate PCA.…”