EmoSesnse-Automatic Speech Recognition Using AIML Implementation Through wav2vec2 Model

The project, titled-“Emosense”, and very much like its name, leverages advanced machine learning algorithms to decode emotional cues embedded within speech patterns. By integrating sentiment analysis, tone detection, and contextual understanding, EmoSense discerns not only the words spoken but also the underlying emotions, empowering businesses to deliver personalised and empathetic responses This project is designed to analyse and classify human emotions from audio recordings of speech. It utilises the RAVDESS dataset, which contains a variety of emotional speeches from different actors. The data processing step involves loading and cleaning the dataset, and then extracting features from the audio files. The program then uses a deep learning model called Wav2Vec2 to convert the audio recordings into numerical representations. These representations are then fed into a classification model called “HubertForSequenceClassification” to classify the emotions. The program uses a tokenization process to convert the audio data into a format that can be fed into the model. The results of the model are visualised using graphs to show the accuracy and loss during the training process. The final accuracy of the model on the test dataset is 93 percent. The program's ability to accurately classify emotions from speech can provide valuable insights for various fields, including mental health, education, and entertainment. It can also be used to develop assistive technology for individuals with communication difficulties. The model's accuracy and ability to generalise to different speakers and accents can be further improved through additional training and fine-tuning. Overall, this program demonstrates the potential of deep learning models in speech emotion recognition and highlights the importance of developing accurate and robust models for real-world applications.