Showing papers by "Anupam Shukla published in 2008"

A novel approach of speaker authentication by fusion of speech and image features using Artificial Neural Networks

Abstract: This paper presents a decision fusion technique for a bimodal biometric verification system that makes use of facial and speech biometrics. This report considers multimodal biometric systems and their applicability to access control, authentication and security applications. We have simulated three Artificial Neural Network (ANN) models: firstly, speaker identification by speech parameters, secondly person identification by image parameters and finally the person authentication by fusion of speech and image feature. All the three ANN models are trained by Back-propagation algorithm.

Multi Neuron Heuristic Search

Abstract: Summary We know the various searching algorithms available today. Searching has become one of the most essential parts of the artificial intelligence algorithms these days. We have so many algorithms like A*, Heuristic Search, Breadth-First Search, Depth First Search, etc. All these are applied to various problems in their own way. We need to predict the most appropriate search technique as the input data is not known. In this paper we present a new searching algorithm. This algorithm works on the principle of applying many neurons (elementary searching units) for working on different data one after the other. Hence as in the case of A* and heuristic search, we do not only select the best current node, but we select a range of nodes from the best to worst. At each iteration various nodes are seen and expanded which have varying heuristic costs. This algorithm would work very well on data in which heuristics change suddenly from very good to bad or vice-versa. We implemented this algorithm and put it on the mazesolving problem, where the heuristic cost was the distance between the nodes to goal point. We saw that the algorithm worked better than any existing algorithm and visited the least number of nodes. This proves the efficiency of the algorithm. We have also shown that this algorithm lies between A* Algorithm and Breadth First Search. Both these algorithms can be reached using this algorithm.