Results 1 -
6 of
6
Multimodal Biometric Systems: Study to Improve Accuracy and Performance
- International Journal of Current Engineering and Technology
, 2014
"... Abstract ..."
Model level fusion of edge histogram descriptors and gabor wavelets for landmine detection with ground penetrating radar
- In IEEE International Geoscience and Remote Sensing Symposium (IGARSS
, 2010
"... Abstract We propose combining heterogeneous sets of features for a continuous hidden Markov model classifier. We use a model level fusion approach and apply it to the problem of landmine detection using ground penetrating radar (GPR). We hypothesize that each signature (mine or non-mine) can be cha ..."
Abstract
-
Cited by 1 (0 self)
- Add to MetaCart
(Show Context)
Abstract We propose combining heterogeneous sets of features for a continuous hidden Markov model classifier. We use a model level fusion approach and apply it to the problem of landmine detection using ground penetrating radar (GPR). We hypothesize that each signature (mine or non-mine) can be characterized better by multiple synchronous sequences that can capture different and complementary salient. Our work is motivated by the fact that mines and clutter objects can have different characteristics depending on the mine type, soil and weather conditions, and burial depth. Thus, ideally different sets of specialized feature extraction mechanisms, may be needed to achieve high detection and low false alarm rates. In order to fuse the different modalities, a multi-stream continuous HMM that includes a stream relevance weighting component is developed. In particular, we modify the probability density function that characterizes the standard continuous HMM to include state and component dependent stream relevance weights. We modify the maximum likelihood based Baum-Welch algorithm and the Minimum Classification Error/Gradient Probabilistic Descent (MCE/GPD) learning algorithms to include stream relevance weights. We generalize their objective functions and derive the necessary conditions to update all model parameters simultaneously. We use the proposed approach to build an HMM classifier that combines two sets of features. The first one, based on edge histogram descriptor, extracts edges in the time domain. The second set of features extracts edges in the frequency domain at multiple scales and orientations. Results on a large collection of GPR alarms show that the proposed model level fusion outperforms the baseline HMM when each feature is used independently and when both features are combined with equal weights.
A Visualization Framework for Team Sports Captured using Multiple Static Cameras
, 2013
"... ..."
(Show Context)
Fusion of Spectrograph and LPC Analysis for Word Recognition: A New Fuzzy Approach
, 2004
"... Word Recognition is generally difficult and imprecise if we use just one method. In this article, data fusion is applied to word recognition by integration of t wo features extracted form human speech: speech spectrograph and time domain features (spectral coefficients). Four different methods are ..."
Abstract
- Add to MetaCart
Word Recognition is generally difficult and imprecise if we use just one method. In this article, data fusion is applied to word recognition by integration of t wo features extracted form human speech: speech spectrograph and time domain features (spectral coefficients). Four different methods are applied to fusion of these features, including weighted averaging, k-means clustering, fuzzy k-means and fuzzy vector quantization. Simulation results show that fusion of time domain and spectrograph features yields more precise and satisfactory results compared to other methods of word recognition that use just one speech feature for word recognition, like FVQ/MLP (fuzzy vector quantization combined with multi-layered perceptron neural network). The importance of this result is prominent if the signal to noise ratio is low.
Speech, Signature and Handwriting Features
"... Abstract—The objective of this work is to develop a multimodal biometric system using speech, signature and handwriting information. Unimodal biometric person authentication systems are initially developed for each of these biometric features. Methods are then explored for integrating them to obtain ..."
Abstract
- Add to MetaCart
(Show Context)
Abstract—The objective of this work is to develop a multimodal biometric system using speech, signature and handwriting information. Unimodal biometric person authentication systems are initially developed for each of these biometric features. Methods are then explored for integrating them to obtain multimodal system. Apart from implementing state-of-the art systems, the major part of the work is on the new explorations at each level with the objective of improving performance and robustness. The latest research indicates multimodal person authentication system is more effective and more challenging. This work demonstrates that the fusion of multiple biometrics helps to minimize the system error rates. As a result, the identification performance is 100 % and verification performances, False Acceptance Rate (FAR) is 0%, and False Rejection Rate (FRR) is 0%. Keywords- Biometrics; Speaker recognition; Signature recognition; Handwriting recognition; Multimodal system.
Error Level Fusion of Multimodal Biometrics
, 2011
"... This paper presents a multimodal biometric system based on error level fusion. Two error level fusion strategies, one involving the Choquet integral and another involving the t-norms are proposed. The first strategy fully exploits the non additive aspect of the integral that accounts for the depende ..."
Abstract
- Add to MetaCart
(Show Context)
This paper presents a multimodal biometric system based on error level fusion. Two error level fusion strategies, one involving the Choquet integral and another involving the t-norms are proposed. The first strategy fully exploits the non additive aspect of the integral that accounts for the dependence or the overlapping information between the error rates FAR’s and FRR’s of each biometric modality under consideration. A hybrid learning algorithm using combination of Particle Swarm Optimization, Bacterial Foraging and Reinforcement learning is developed to learn the fuzzy densities and the interaction factor. The second strategy employs t-norms that require no learning. The fusion of the error rates using t-norms is not only fast but results in very good performance. This sort of fusion is a kind of decision level fusion as the error rates are derived from the decisions made on individual modalities. The experimental evaluation on two hand based datasets and two publically available datasets confirms the utility of the error level fusion.
