Intelligent Processing Algorithms for Motion Capture Data Oriented to Artificial Intelligence

Traditional motion capture systems are prone to environmental interference, resulting in noise and errors in the captured data. This article proposes an artificial intelligence oriented intelligent processing algorithm for motion capture data, which improves the quality of motion capture data and reduces noise and errors. This article utilizes sensors to collect a large amount of motion timing data and motion image data, preprocesses the data, and constructs a Convolutional Neural Network (CNN) Long Short Term Memory (LSTM) model. By using the CNN model to extract features from images, this paper uses the LSTM model to analyze temporal data, fuses the output of LSTM and CNN, and finally estimates the motion attitude. The data results of the test set indicate that the average mean square error of motion attitude estimation using the CNN-LSTM model is only 0.03. The use of the CNN-LSTM model can effectively improve the performance of motion data pose analysis.