Publication

Abstract : Modern engineering society requires qualified solutions to many practical problems, to meet accurate outcomes. Emotion detection through face landmarks has unique coordination toward improving the feedback system. Bio-inspired learning methods are an interesting area of research. Face expressions are detected through various methodologies. Bio-inspired techniques are used to extract facial landmarks to detect facial expressions even better. The selection of face attributes is crucial for the prediction of deeper emotions. The proposed system is focused on implementing a bio-inspired facial landmark detection system for emotion extraction. The dataset used for the analysis is taken from the (JAFEE) Japanese female face expressions dataset. It consists of a collection of unique expressions and recorded images. Human face attributes are extracted from the Haar cascade model. The proposed novel methodology is derived from the Robust Emo-Spot Extracting technique (RESET) using point mapping model-based landmark mapping for pre-trained images and labeling spots. The face images are correlated with databases and classified the type using a deep convolutional neural network, in which appropriate Emo-spots are mapped using RESET. The adjustable Emo-Spots are randomly tuned using the randomly looped K-nearest neighbor algorithm. The novel algorithm checks for the correlation ratio (CR) <20. Non-correlated values result in higher values of CR >20. The proposed system is compared with the state-of-art approach in terms of accuracy.