A Compressive Sensing Scheme of Frequency Sparse Signals for Mobile and Wearable Platforms

In selected scenarios, sensor data capturing with mobile devices can be separated from the data processing step. In these cases, Compressive Sensing allows a significant reduction of the average sampling rate below the Nyquist rate, if the signal has a sparse frequency representation. This can be motivated in order to increase the energy efficiency of the mobile device and extend its runtime. Since many signals, especially in the field of motion recognition, are time-dependent, we propose a corresponding general sampling algorithm for time-dependent signals. It even allows a declining average sampling rate if the data acquisition is extended beyond a projected acquisition end. The presented approach is testified for the purpose of motion recognition by evaluating real acceleration sensor data acquired with the proposed algorithm.