Research

Research Interests

  • Computational Forensics: Alteration detection, writer recognition, signature verification, speaker recognition

  • Multimedia Forensics: Audio/video/image authentication, device recognition

  • Image Processing: Feature Extraction and Analysis

  • Pattern Recognition: Learning and Classification

Device Recognition from audio recordings

In this study, we explored the effect of format change in device recognition from audio recordings. The study was conducted on 21 mobile phones of various makes and models and nine different file formats both lossy and lossless in nature. The device was modeled using Mel Frequency Cepstral Coefficient (MFCC) with Gaussian Mixture Model framework. No significant difference in accuracy was found across the file formats.

Sparse Model for Forensic Writer Recognition

In this study, we propose a novel sparse model for writer recognition. It is based on the assumption that an individual writer generates a particular kind of structural primitives (graphemes) and the writing habits of that particular writer can be represented as the linear combination of basis vectors of such graphemes

Alteration Detection Pattern Recognition Framework

A pioneer work on the detection of alteration in the check due to difference of ink was explored in pattern recognition framework. Color features were extracted from trichromatic and opponent chromaticity space. Texture features were extracted using GLCM. Finally, detection of alteration was done using KNN, MLP and SVM classifiers