Background: The compressed sensing (CS) of acceleration data has been drawing\nincreasing attention in gait telemonitoring application. In such application, there\nstill exist some challenging issues including high energy consumption of body-worn\ndevice for acceleration data acquisition and the poor reconstruction performance due\nto nonsparsity of acceleration data. Thus, the novel scheme of compressive sensing of\nacceleration data is needed urgently for solutions that are found to these issues.\nMethods: In our scheme, the sparse binary matrix is firstly designed as an optimal\nmeasurement matrix only containing a smallest number of nonzero entries. And then\nthe block sparse Bayesian learning (BSBL) algorithm is introduced to reconstruct acceleration\ndata with high fidelity by exploiting block sparsity. Finally, some commonly\nused gait classification models such as multilayer perceptron (MLP), support vector\nmachine (SVM) and KStar are applied to further validate the feasibility of our scheme\nfor gait telemonitoring application.\nResults: The acceleration data were selected from open Human Activity Dataset of\nSouthern California University (USC-HAD). The optimal sparse binary matrix (a smallest\nnumber of nonzero entries is 8) is as strong as the full optimal measurement matrix\nsuch as Gaussian random matrix. Moreover, BSBL algorithm significantly outperforms\nexisting conventional CS reconstruction algorithms, and reaches the maximal signalto-\nnoise ratio value (70 dB). In comparison, MLP is best for gait classification, and it can\nclassify upstairs and downstairs patterns with best accuracy of 95 % and seven gait\npatterns with maximal accuracy of 92 %, respectively.\nConclusions: These results show that sparse binary matrix and BSBL algorithm are\nfeasibly applied in compressive sensing of acceleration data to achieve the perfect\ncompression and reconstruction performance, which has a great potential for gait\ntelemonitoring application.
Loading....