Generation of Dance Movement Sequences Based on Audio Information

Abstract

The task of action sequence generation based on audio is a cross-modal generation task, which automatically generates continuous action sequences with similar or consistent time, semantics or emotion with the input information through the input audio signal. With advances in computer vision, as well as digital entertainment, methods that link human speech to digital body movements have made rapid progress. At present, the main methods are based on neural networks and deep learning for multi-modal generation. Through the multi-modal generation method based on computer network, it often faces the problems of low correlation between the generated content and the input content, low overall fluency, and unclear emotional expression. Based on the systematic review of the existing literature, this paper analyzes the mainstream methods in the task of dance movement generation. Finally, the key problems to be solved in this field are discussed, and the future research content and direction are prospected.