,
PhD Research Scholar, Department of Computer Science Engineering, Koneru Lakshmaiah Education Foundation , Vaddeswaram, , India
,
Associate Professor, Department of CSE, Koneru Lakshmaiah Education Foundation , Vaddeswaram , India
Associate Professor, Department of CSE, GRIET , Hyderabad , India
Micro-expressions (MEs) are involuntary facial expressions, short-lived (usually between 1/5 and 1/25 seconds), and important in the application of security, psychological tests, and forensics. The MEs are however difficult to identify because it occur quickly and also involve little movement of the muscles. The paper presents an Enhanced Micro-Expression Descriptor which incorporates Gabor Wavelet-based Local Binary Patterns (GW-LBP), Temporal Gradient Magnitude Histograms (TGMH), and Wavelet Transform (WT) to enhance ME recognition, by overcoming the weaknesses of traditional methods in illumination sensitivity and poor computing power. The algorithm involves the use of GW-LBP to extract spatial texture, TGMH to capture changes in temporal motion, and WT to analyze frequencies on a multiscale basis. This is achieved by classifying the fused feature set with an RBF kernel Support Vector Machine (SVM), which is optimized by down-sampling to a size manageable by resources (4096 dimensions) to provide a resource-efficient, real-time solution with application in edge computing. Benchmark dataset experimental results prove that the proposed method is better than the existing techniques with a recognition accuracy of 85.9%. This is a major boost compared to conventional procedures such as the LBP-TOP (67.5%) and CNN-based models (78.3%). Also, the Wavelet Transform option exploited the highest score in entropy (0.93), which implies that it can be highly used in real-time behavioral analysis, emotion detection, and security surveillance. The findings affirm that the hybrid approach, which incorporates spatial, temporal, and frequency characteristics, has a better performance than the existing ME recognition models.
This is an open access article distributed under the Creative Commons Attribution Non-Commercial License (CC BY-NC) License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
0
The statements, opinions and data contained in the journal are solely those of the individual authors and contributors and not of the publisher and the editor(s). We stay neutral with regard to jurisdictional claims in published maps and institutional affiliations.