In order to improve the performance of joint reconstruction of multi-sensor acceleration data from different wearable devices, a novel approach to jointly reconstruct based on distributed compressed sensing (DCS) algorithm was proposed. The basic idea was that the raw data was firstly compressed through encoding, and the encoded data was sent to remote terminal. Then, with the spatiotemporal correlation of data from sensors, the joint reconstruction method based on Block Sparse Bayesian Learning (BSBL) was applied to decode the compressed data at remote terminal. At last, the wearable data from University of California-Berkeley database was analized. Experiments show that the proposed approach can gain better performance than the traditional joint reconstruction algorithms such as TMSBL and tMFOCUSS, and decode the compressed data accurately. The proposed technique may be helpful for telemedicine application.