Two-step Wavestrapping: Simulating Non-stationary Acceleration Data in the Mobile Computing Context

Measurements of acceleration provide an important source of information in understanding and predicting the movements of mobile computer users. However, collecting acceleration data can require substantial time and resources because it involves both human participation and the construction of adequate contextual environments. Thus, simulating acceleration data would be helpful to circumvent these difficulties. However, using traditional resampling techniques turned out to be inadequate for simulation of these non-stationary time series data. The present study proposes a wavelet-based resampling approach, called the “two-step wavestrapping,” which consists of parallel wavestrapping and energy/trend adjustments. This approach was applied to the acceleration data we collected from mobile computing users under six different contextual settings. The results showed that two-step wavestrapping can successfully generate surrogate acceleration data from the collected acceleration data.