Adaptive compressed sensing for acceleration data transmission in human motion capture

In a human motion capture system, massive acceleration data will cost much transmission time. In order to reduce the acceleration data transmission volume while guaranteeing measurement accuracy, this paper presents an adaptive compressed sensing algorithm for acceleration data transmission in human motion capture. An innovative feature of our approach is to associate the sparsity and best Compression Ratio of acceleration data in a learning phase. Based on this correlation, the sensor node uses sparsity to describe the structural characteristics of acceleration data, and it adjusts best Compression Ratios of data according to the characteristics of the real-time acceleration data. The results of experiment show that algorithm can adjust the best Compression Ratios of original acceleration data at different parts of body in several typical motions, while ensuring Reconstruction Accuracy. For example, in running, the approach can achieve that the Compression Ratios at arms and legs separately reach to approximately 80% and 60%, while the compression data can achieve the same Reconstruction Accuracy. What's more, the algorithm can increase the Compression Ratio of capture system by about 20% compared with the current BSBL-BO algorithm, while achieving the same minimum Reconstruction Accuracy of capture system.