Motion capture data analysis based on neural network

With the development of technology, the application of human motion prediction technology is becoming more and more extensive. In the virtual world, human motion prediction plays an increasingly important role in the fields of film action synthesis, animation production and game character control, which makes the movement of virtual characters more vivid and closer to the movement of real human beings. At the same time, human motion prediction technology is also combined with hardware equipment in the real world, which provides a new application possibility for human rehabilitation auxiliary medical equipment. At present, the mainstream methods of human motion prediction model construction rely on huge motion database and state machine tween technology. In the process of model construction, the key frames are matched and selected from the action database, and the key frames are smoothly connected through the state machine tween technology. However, this production method is complex and expensive, and the degree of reproduction of the predicted human motion depends largely on the comprehensiveness of the motion database and the naturalness of the state machine tween. Therefore, the development of human motion prediction technology is of great significance for improving animation, games in the virtual world and rehabilitation auxiliary medical equipment in the real world. In order to solve the above problems, this paper proposes the motion capture data analysis based on neural network. The attention mechanism is added to the time convolutional network, and the ATTLS double residual linkage model is constructed. The introduction of multi-head self-attention model into this model can better assign different weights to the data at different moments according to the internal relationships of the time series data, thus contributing to the data prediction in the next step. Finally, the reliability of the model is verified by experiments.