Toward a proposed framework for mood recognition using LSTM Recurrent Neuron Network

Abstract: Affective analysis plays an important role in understanding human characteristics, predicting human behavior and diagnosing mental health problems. Although a large number of affective recognition researches have been published, predicting mood - the long lasting affect from real life context is still a challenge because of the complexity of correlations between mood and daily factors. We therefore aim at developing a framework for predicting mood considering several aspects on human and environment adaption. The framework consists of two components: a data acquisition tool designed for wearable devices and a mood model based on Long Short-Term Memory Recurrent Neuron Network (LSTM-RNN). In this paper, we are focused on presenting the later component to apply LSTM RNN for mood prediction in daily life.