A Deep Convolutional Neural Network Based Virtual Elderly Companion Agent

Abstract

This study presents a Virtual Elderly Companion Agent that based on speech spectrograms and deep convolutional neural networks. The system can dynamically detect and analyze the user's emotion from the dialogue and give appropriate positive feedback. The proposed system architecture is divided into two parts. The client side supports Android operating system; the server side is implemented in python, and applied GoogleLeNet and AlexNet for emotion recognition. The system supports natural language speech input, and then analyzes the converted speech spectrogram to provide appropriate feedback.