Integrating Bi-LSTM with Deep Learning for Emotion Analysis from Brain Wave Patterns

Emotion detection and classification are becoming new frontiers in the capability of brain machine interface in the present times. Several studies have demonstrated that audio, video, as well as EEG data are helpful in the automated detection of affective states. EEG based affect detection is an important aspect of psychiatric evaluation for persons. In case sensor EEG data is gathered from various experimental sessions or participants, the signals obtained tend to be non-stationary. Building an intelligent emotion recognition system, which performs fair in terms of accuracy remains a problem due to the presence of numerous noisy, non-stationary, and nonlinear EEG signals. Some researchers have provided evidence that specific EEG brain waves reflect certain thoughts or feelings. This research proposes a completely new system responsible for automatically recognizing the emotions of users using deep learning techniques of computer network EEG signals acquired during playing computer games. EEG data from 28 participants were acquired on 14 channels Emotive Epoc + portable and wearable devices. Each participant played four distinct emotional games for five minutes each, yielding a total of 20 minutes of EEG data per volunteer. It is also shown that this framework is capable of identifying four fundamental emotion types in real time during games. The findings indicate that the suggested model-based emotion identification framework is a useful way for detecting emotions from EEG data. The network achieves 99.32% accuracy while using less computation time.