Mouth Gesture Classification using Computational Intelligence

Computational intelligence works as a recent advancement in many fields of classifications and predictions. The mouth detection approach classifies images based on the value of simple visual attributes and has been shown to be reliable under a variety of lighting situations. The method uses a cataract of enhanced classifiers with simple Haar-wavelet-like features on various sizes and positions. Mouth images are recognised using the Haar-cascaded technique, HOG features are extracted from the detected regions, modelled with Support Vector Machine, and LSTM classified into mouth shut, mouth open, tongue left, and tongue right in the suggested work. According to the findings, HOG characteristics combined with SVM and LSTM accurately locate the mouth regions. The models are compared to see how well they perform. SVM and LSTM gives the best results, with a 88 percent accuracy rate.