Saurabh Bhardwaj

GFM-Based Methods for Speaker Identification

This paper presents three novel methods for speaker identification of which two methods utilize both the continuous density hidden Markov model (HMM) and the generalized fuzzy model (GFM), which has the advantages of both Mamdani and Takagi-Sugeno models. In the first method, the HMM is utilized for the extraction of shape-based batch feature vector that is fitted with the GFM to identify the speaker. On the other hand, the second method makes use of the Gaussian mixture model (GMM) and the GFM for the identification of speakers. Finally, the third method has been inspired by the way humans cash in on the mutual acquaintances while identifying a speaker. To see the validity of the proposed models [HMM-GFM, GMM-GFM, and HMM-GFM (fusion)] in a real-life scenario, they are tested on VoxForge speech corpus and on the subset of the 2003 National Institute of Standards and Technology evaluation data set. These models are also evaluated on the corrupted VoxForge speech corpus by mixing with different types of noisy signals at different values of signal-to-noise ratios, and their performance is found superior to that of the well-known models.

A novel shape based batching and prediction approach for time series using HMMs and FISs

This paper introduces a novel approach which uses a Hidden Markov Model (HMM) based Fuzzy Inference System (FIS) for prediction of systems that are non deterministic, dynamical and chaotic in nature. The HMM is used for shape based batch creation of training data which is then processed one batch at a time by a FIS. The Membership functions and Rule Base of the FIS are tweaked to predict the correct output for an input dataset. The novel Prediction method used here exploits the Pattern Identification prowess of the HMM for batch selection and the FIS of each batch to predict the output of the system. The Benchmark applications of the Mackey Glass Time Series (MGTS) as well as the Sunspot Data time-series were used for testing the competence of this method.

Chaotic time series prediction using combination of Hidden Markov Model and Neural Nets

This paper introduces a novel method for the prediction of chaotic time series using a combination of Hidden Markov Model (HMM) and Neural Network (NN). In this paper, an algorithm is proposed wherein an HMM, which is a doubly embedded stochastic process, is used for the shape based clustering of data. These data clusters are trained individually with Neural Network. The novel prediction approach used here exploits the Pattern Identification prowess of the HMM for cluster selection and uses the NN associated with each cluster to predict the output of the system. The effectiveness of the method is evaluated by using the benchmark chaotic time series: Mackey Glass Time Series (MGTS). Simulation results show that the given method provides a better prediction performance in comparison to previous methods.

Wavelet Packet Based Mel Frequency Cepstral Features for Text Independent Speaker Identification

The present research proposes a paradigm which combines the Wavelet Packet Transform (WPT) with the distinguished Mel Frequency Cepstral Coefficients (MFCC) for extraction of speech feature vectors in the task of text independent speaker identification. The proposed technique overcomes the single resolution limitation of MFCC by incorporating the multi resolution analysis offered by WPT. To check the accuracy of the proposed paradigm in the real life scenario, it is tested on the speaker database by using Hidden Markov Model (HMM) and Gaussian Mixture Model (GMM) as classifiers and their relative performance for identification purpose is compared. The identification results of the MFCC features and the Wavelet Packet based Mel Frequency Cepstral (WP-MFC) Features are compared to validate the efficiency of the proposed paradigm. Accuracy as high as 100% was achieved in some cases using WP-MFC Features.

A novel hybrid model for solar radiation prediction

Due to the increasing requirement for the design, optimization and performance evaluation of the solar energy systems modeling and prediction of solar radiation is of considerable importance. A number of solar radiation prediction models have been developed, ranging from simple empirical models to very complicated computer coded models. This paper presents Generalized Fuzzy Model (GFM) based paradigm for the prediction of solar radiation. The present research uses both the probability theory and fuzzy set theory for the prediction of radiation. For collecting the meteorological parameters a test bed is made at Solar Energy Centre, Gurgaon, India. Both the models are applied on four different sets comprising of various combination of input meteorological parameters (Day of the year, sun shine hour, ambient temperature, relative humidity and atmospheric pressure). The results of prediction are encouraging.

A novel shape based batching and prediction approach for sunspot data using HMMs and ANNs

This paper introduces a novel approach which uses a Hidden Markov Model (HMM) based Artificial Neural Networks (ANN) for prediction of systems that are non deterministic, dynamical and chaotic in nature. The HMM is used for shape based batch creation of training data, which is then processed one batch at a time by an ANN. The weights and Learning Rate of the ANN are tweaked to predict the correct output for an input dataset. The novel Prediction method used here exploits the Pattern Identification prowess of the HMM for batch selection and the ANNs of each batch to predict the output of the system. The Standard application of the Sun-Spot Data (SSD) was used for testing the competence of this method.

Multi-level Fusion of Palmprint and Dorsal Hand Vein

A novel multilevel level fusion of palmprint and dorsal hand vein is developed in this work. First feature level fusion is done on left and right hand palmprint to get feature fused vector (FFV). Next, the scores of FFV and veins are calculated and score level fusion is done in order to identify person. Hence both the feature level as well as score level fusion techniques have been used in a hybrid fashion. In the present work, feature fusion rules have been proposed to control the dimension of FFV. For palmprint, IIT Delhi Palmprint Image Database version 1.0 is used which has been acquired using completely touchless imaging setup. In this feature level fusion of both left and right hand is used. For dorsal hand veins, Bosphorus Hand Vein Database is used because of the stability and uniqueness of hand vein patterns. The improvement of results verify the success of our approach of multilevel level fusion.

Pattern-Similarity-Based Model for Time Series Prediction

This research proposes a pattern/shape‐similarity‐based clustering approach for time series prediction. This article uses single hidden Markov model (HMM) for clustering and combines it with soft computing techniques (fuzzy inference system/artificial neural network) for the prediction of time series. Instead of using distance function as an index of similarity, here shape/pattern of the sequence is used as the similarity index for clustering, which overcomes few of the shortcomings associated with distance‐based clustering approaches. Underlying hidden properties of time series are captured with the help of HMM. The prediction method used here exploits the pattern identification prowess of the HMM for cluster selection and the generalization and nonlinear modeling capabilities of soft computing methods to predict the output of the system. To see the validity of the proposed method in the real‐life scenario, it is tested on four different time series. The first is a benchmark Mackey–Glass time series, which is tested for delay parameters τ = 17 and τ = 30. The remaining time series are monthly sunspot data time series, Laser data time series and the last is Lorenz attractor time series. Simulation results show that the proposed method provide a better prediction performance in comparison with the existing methods.

A Novel Clustering Approach Using Shape Based Similarity

The present research proposes a paradigm for the clustering of data in which no prior knowledge about the number of clusters is required. Here shape based similarity is used as an index of similarity for clustering. The paper exploits the pattern identification prowess of Hidden Markov Model (HMM) and overcomes few of the problems associated with distance based clustering approaches. In the present research partitioning of data into clusters is done in two steps. In the first step HMM is used for finding the number of clusters then in the second step data is classified into the clusters according to their shape similarity. Experimental results on synthetic datasets and on the Iris dataset show that the proposed algorithm outperforms few commonly used clustering algorithm.

Identification of Hidden Information Using Brain Signals

A lot of information can be extracted and understood from brain signals. The brain gives an insight into the various processes underlying our behavior and responses. If an individual is hiding information in the brain, this information can be detected using brain signals. In the present work, electroencephalogram (EEG) signals of the subjects are analyzed to detect hidden information in the individuals. After preprocessing the data, wavelet features are extracted. Initially, classification algorithm based on support vector machine (SVM) has been used to identify the subjects hiding the information. Afterwards, k-nearest neighbor (kNN) algorithm has been used to distinguish between the guilty and innocent participants. SVM performs better than kNN and achieves a significant classification accuracy of 80% in identification of the concealed information.