A Novel Model for Emotion Detection with Multilayer Perceptron Neural Network
M. Muthumari, V Akash, K. PrudhviCharan, P. Akhil · 2022 6th International Conference on Intelligent Computing and Control Systems (ICICCS) · 2022
Identifying emotion from speech is a very important concept in the field of Human Computer Interaction (HCI), and it has been gaining progressive interest in the current research area. As a result, various researchers have introduced and established numerous systems and methods for recognising emotion in human speech, which discussed earlier implementations and advances in emotion recognition. The goal of creating this model is to detect sounds in audio files. This proposed system also includes speech emotion detection, which can detect emotions such as Sad, Anger, and Happiness in audio signals, gender detection and playing an YouTube video according to the detected emotion. Whereas in prior projects, just emotion detection techniques were used, however here, emotion as well as gender are detected. This output is sent as an input to YouTube, which will play the music based on the user's mood, causing the user's mood to settle quickly. The CNN feature extraction method is used, the feature vector dimensions are handled by using NumPy and the audio classification is based on MFCC (Mel frequency cepstral coefficients). To train and test the model, two datasets: RAVDESS and SAVEE, are primarily used and a new deep learning model is created and trained using the collected datasets. The model is then saved, and loaded with the inputs so that, the required results can be retrieved. The system scope is the device on which the Jupiter notebook software is installed, allowing to run the code.