A Proposed Technique For Recognizing And Analyzing Body Language Using Slantlet Transform And Deep Learning

Body language encompasses a variety of nonverbal cues, such as head movements, hand movements, body postures, or any other part of the body, from which intentions and feelings can be inferred. The expanding use of body language in various fields has made it necessary for recognition systems to achieve accuracy and efficiency in performance. This paper proposes a two-stage to body language recognition and analysis technique. The first stage is extracting frames from the video, after which the third level of Slantlet Transform is applied to reduce the dimensions of the frames while preserving the essential features and removing unnecessary features, thus reducing processing time. In the second stage, deep learning models are used to extract features and classify, followed by analyzing body language and human behavior in various legal contexts. This mechanism helps reduce processing time while maintaining high accuracy. The Slantlet Transform also makes the technique resistant to rotation and noise, as it removes unwanted features that typically contain noise. A variety of experiments were conducted to test the accuracy and efficiency of the proposed technique using the GEMEP dataset and The Real-Life Deception Detection dataset. The proposed technique achieved an accuracy of 99.56% and 98.22% in recognizing and analyzing body language, respectively, in addition to achieving a reduction in processing time of more than 75%.