Non-verbal Cues Detection from Depth Video with Patterns of Oriented Motion Flow–An Indispensable Tool for Guilt Investigation

K. Sathish, Rishma, Dolly Shah, Shagun Goyal and Prahitya Mahavi

We see criminal investigation that are carried out by the investigation agencies use psychologists to identify the guilty and to know if someone is lying. There have been many system that use facial recognition system to identify emotions. The Pattern of oriented motion flow (POMF) is one of them in the system proposed earlier there is no way to detect the body language changes. This system proposes a use case of criminal investigating where in patterns of oriented motion flow (POMF) is used to detect the subtle non verbal changes in a convict by creating a baseline of changes and detecting erratic and sporadic changes based on the velocity of reaction to any stimuli during the investigation process. A depth sensing camera is used to record the person and every frame fed into the neural network. Based on orientation changes in horizontal and vertical axes between every frame and the velocity at which it changes, the directional vectors are plotted on a POMF histogram which is fed into a Hidden Markov chain and results are predicted by using K-Means Clustering to spot the cue of emotions behind the body language of the suspect with accuracy.

Volume 11 | 04-Special Issue

Pages: 1138-1144