Deception Discernment using Audio and Micro-expressions

This work presents an innovative method for deception discernment by combining facial action units (AUs) with speech analysis utilizing Mel-frequency cepstral coefficients (MFCC) to identify microexpressions. To improve the precision of deception detection, the suggested model fuses data from both modalities using a Long Short-Term Memory (LSTM) neural network. Our method, serving as alternative conventional polygraph testing, collects minute cues in facial expressions and speech to study the pattern of deception discernment. By utilizing the temporal dynamics present in human communication, the synergistic coupling of MFCC and AUs improves the model's sensitivity and specificity in identifying misleading intent. This study is a promising inclusion in the study of deception detection.