Detection of Human Emotion Recognition using Hidden Markov Model and Bidirectional Associative Memory with voice

Abstract

[Detection of Human Emotion Recognition using Hidden Markov Model and Bidirectional Associative Memory with voice] A person's emotional control system through voice can use the Hidden Markov Model (HMM) algorithm. However, to see if the performance of the HMM algorithm in the application system is optimal or not, a comparison is needed in order to obtain maximum results. Therefore, the researchers performed the performance of recognizing a person's emotions using the HMM algorithm and the Bidirectional Associative Memory (BAM) algorithm through voice. The Hidden Markov Model (HMM) consists of a Markov chain in the first part that hides the state, therefore the internal behavior of the model remains invisible. While the BAM algorithm can process incomplete input, because of the reciprocal relationship between the output layer to the input layer. In the BAM algorithm, the value of the test sound and the value of the training voice sample obtained will be searched for the vector value using a weight value search which is done by changing the binary matrix into a bipolar matrix. In this study, we will create an application system that can detect sounds in the form of angry, happy, and neutral emotions. And the database used is the sound of film recordings. This research was conducted to produce a system that can recognize the probability of emotions in the angry, happy and neutral categories, namely by showing the performance of the two methods so that we can find out which method produces the maximum output.