Pawan K. Ajmera

Text-independent speaker identification using Radon and discrete cosine transforms based features from speech spectrogram

This paper presents a new feature extraction technique for speaker recognition using Radon transform (RT) and discrete cosine transform (DCT). The spectrogram is compact, efficient in representation and carries information about acoustic features in the form of pattern. In the proposed method, speaker specific features have been extracted by applying image processing techniques to the pattern available in the spectrogram. Radon transform has been used to derive the effective acoustic features from the speech spectrogram. Radon transform adds up the pixel values in the given image along a straight line in a particular direction and at a specific displacement. The proposed technique computes Radon projections for seven orientations and captures the acoustic characteristics of the spectrogram. DCT applied on Radon projections yields low dimensional feature vector. The technique is computationally efficient, text-independent, robust to session variations and insensitive to additive noise. The performance of the proposed algorithm has been evaluated using the Texas Instruments and Massachusetts Institute of Technology (TIMIT) and our own created Shri Guru Gobind Singhji (SGGS) databases. The recognition rate of the proposed algorithm on TIMIT database (consisting of 630 speakers) is 96.69% and for SGGS database (consisting of 151 speakers) is 98.41%. These results highlight the superiority of the proposed method over some of the existing algorithms.

Fractional Fourier transform based features for speaker recognition using support vector machine

This paper presents a text-independent speaker recognition technique in which the conventional Fourier transform in Mel-Frequency Cepstral Coefficient (MFCC) front-end is substituted by fractional Fourier transform. Support Vector Machine (SVM) maps these input features into a high-dimensional space to separate classes by a hyperplane with enhanced discrimination capability. SVM based on mean-squared error classifier produces more accurate system. The Fractional Fourier Transform (FrFT) reveals the mixed time and frequency components of the signal. Modelling of speech signals as mixed time and frequency signals represents better production and perception speech characteristics. Processing of time-varying signals in fractional Fourier domain allows us to estimate the signal with least Mean Square Error (MSE) making the technique robust against additive noise compared to Fourier domain maintaining same computational complexity. The feasibility of the proposed technique has been tested experimentally using Texas Instruments and Massachusetts Institute of Technology (TIMIT) and Shri Guru Gobind Singhji (SGGS) databases. The experimental results show the superiority of the proposed method.

Multiresolution Feature Based Subspace Analysis for Fingerprint Recognition

The image intensity surface in an ideal fingerprint image contains a limited range of spatial frequencies, and mutually distinct textures differ significantly in their dominant frequencies. This paper presents a multiresolution feature based subspace technique for fingerprint recognition. The technique computes the core point of fingerprint and crops the image to predefined size. The multiresolution features of aligned fingerprint are computed using 2-D discrete wavelet transform. LL component in wavelet decomposition is concatenated to form the fingerprint feature. Principal component analysis is performed on these features to extract the features with reduced dimensionality. The algorithm is effective and efficient in extracting the features. It is also robust to noise. Experimental results using the FVC2002 and Bologna databases show the feasibility of the proposed method..

Speaker Recognition Using Auditory Features and Polynomial Classifier

This paper presents a speaker recognition method which makes use of auditory features and polynomial classifier for speaker recognition. Auditory features based on an auditory periphery model extract significant speaker characteristics. Polynomial classifier has been used to accomplish speaker recognition task. Polynomial classifier has several advantages over the conventional classifiers such as computational scalability with the number of speakers, discriminative training allowing it to use out of class data and the statistical interpretation of scoring allowing it to combine with HMM and GMM. This approach achieves substantial performance improvement in a speaker identification task compared with state-of-the-art in a wide range of signal to noise conditions.

Adaptive artificial neural network based control strategy for shunt active power filter

Shunt active power filter (SAPF) is used to mitigate the current harmonics and to improve the power factor. In this paper, Adaptive linear-neuron (ADALINE) based phase lock loop (PLL) controlling scheme is presented for SAPF. ADALINE networks estimate the fundamental supply frequency. This scheme detects the phase information of the supply voltage and also used for parallel computing as it provides faster convergence. This algorithm is trained by least-mean squares (LMS) rule which offers low computational burden on the system. In this work, ADALINE is tuned using particle swarm optimization (PSO) technique to improve the dynamic performance of the system. The results obtained are compared with conventional PLL control technique and are found to be significantly better. The performance of the proposed ADALINE based control algorithm is validated using MATLAB/Simulink.

A multi-modal and multi-algorithmic biometric system combining iris and face

In this paper, we have developed an algorithm which combines features from human iris and face for person verification. Iris recognition is one of the most accurate biometric modalities having verification results close to 98%. On the other hand, face is one of the most widely used biometric features because of its ease of capture. We have adapted score level fusion strategy for our system. However, in addition to this, we are using two different features for face: Gabor filters based and Local Binary Patterns (LBP) based. The iris features are extracted using Daugmans Gabor filters based approach. Using this information, we have developed a multi-modal (combining iris and face), multi-algorithmic (using two different algorithms for feature extraction from face) biometric system. With this system, we achieved more than 85% improvement in the verification performance in terms of Equal Error Rate as compared to the uni-biometrics based system.

Real time human face location and recognition system using single training image per person

This paper presents an automatic real time face location and recognition system. The proposed approach detects the face using the combination of hue, saturation and intensity (H SI) and luminance, red chrominance and blue chrominance (Y CrCb) color Space models. The left most, right most and top most pixels of face are detected using threshold values of parameters. One of the eyes is located using the blue chrominance. The second eye, center of the eyes, and the bottom most part of face is detected using geometrical similarity. The face is cropped using these defined boundaries to extract facial region only. The facial features of cropped image are extracted using the combination of Radon and wavelet transform. The technique computes Radon projections in different orientations and captures the directional features of face images. Further the wavelet transform applied on Radon space provides multiresolution features of the facial images. For classification, the nearest neighbor classifier has been used. The performance and robustness of the proposed system is tested on a face database of 785 images of 157 subjects acquired in conditions similar to those encountered in real world applications. The system achieves a recognition rate of 97.8 % and an equal error rate (EER) of about 2.4% for 157 subjects.

Temperature compensation of ISFET based pH sensor using artificial neural networks

This paper presents a new Machine Learning based temperature compensation technique for Ion-Sensitive Field-Effect Transistor (ISFET). The circuit models for various electronic devices like MOSFET are available in commercial Technology Computer Aided Design (TCAD) tools such as LT-SPICE but no built-in model exists for ISFET. Considering SiO2 as the sensing film, an ISFET circuit model was created in LT-SPICE and simulations were carried out to obtain characteristic curves for SiO2 based ISFET. A Machine Learning (ML) model was trained using the data collected from the simulations performed using the ISFET macromodel in the read-out circuitry. The simulations were performed at various temperatures and the temperature drift behavior of ISFET was fed into the ML model. Constant pH (predicted by the system) curves were obtained when the device is tested for various pH (7 and 10) solutions at different ambient temperatures.

Robust feature extraction from spectrum estimated using bispectrum for Isolated Word Recognition

Extraction of robust features from noisy speech signals is one of the challenging problems in Automatic Speech Recognition (ASR). For Gaussian process, its bispectrum and all higher order spectra are identically zero, which means that bispectrum removes the additive white Gaussian noise while preserving the magnitude and phase information of original signal. Using this bispectrum property, spectrum of original signal can be recovered from its noisy version. Robust Mel Frequency Cepstral Coefficients (MFCC) are extracted from the estimated spectral magnitude (denoted as Bispectral-MFCC (BMFCC)). The effectiveness of BMFCC has been tested on TI-46 isolated word database in noisy (additive white Gaussian) environment. The experimental results show the superiority of the proposed technique over conventional methods for Isolated Word Recognition (IWR).