Standard Time Estimation of Manual Tasks via Similarity Measure of Unequal Scale Time Series

The analysis of standard cycle times for manual tasks has been an important subject in time and motion studies for developing a standardized work process for which the laborious and continuous observation of tasks using a time measurement instrument was usually required. In order to automate this procedure, a motion recognition method is proposed to identify the precise start and end times of manual tasks. To do this, we consider the time series of the hand posture and movement data acquired by a depth-sensing camera. The pattern of motions made to complete a single task is represented by the sign sequence of wavelet coefficients. We then extract the start and end times of each individual task from the original time series of repetitive manual tasks; this is done by searching a set of subtime series of unequal scale that has a similar sign sequence as the prespecified reference. The performance of the proposed procedure is statistically examined by a paired t-test at significance level α = 0.05$ in comparison with a conventional video playback analysis. The mean absolute percentage gap between the estimated standard time and the actual operation time varies from 1.07% to 7.17%.

I. Introduction

Manual tasks performed by the same person at different times represent a nonstationary periodicity due to irregular human behaviors [1]–[3] and unavoidable constraints encountered during the job operations [4], even if the person performs the tasks several times following a predefined standard process. Hence, a motion analysis for determining the standard cycle times for repetitive tasks has been an important subject in several disciplines such as a productivity analysis in time and motion study [5], [6], a periodic activity analysis in sports science [7], video surveillance for homeland security applications [8], and the motif discovery of a time series for healthcare systems [9].