Spectral analysis of upper limb kinematic measurements has been previously demonstrated useful for quantifying physical stress in repetitive motion tasks. The method requires manually separating the data into segments corresponding to individual tasks or work elements, computing power spectra, and averaging. This study investigated using signal pattern recognition to help further automate the analysis by separating the data through identification of stereotypical patterns for cyclical tasks. Joint angles for five industrial jobs were continuously measured using electrogoniometers attached to the wrist, elbow and shoulder of the dominant limb. A multimedia computer system enabled the analyst to review the videotape and interactively indicate element breakpoints. The breakpoints were also automatically identified using a template matching (TM) algorithm. The algorithm identified the cycle terminal points within 0.997 s (S.D.=2.762 s) of the human analyst's reference breakpoint. Analysis using TM breakpoint identification resulted in average differences in the spectra of 9.3° for RMS joint deviation, 14.1° for mean joint angle, and 0.445 Hz for repetition frequency. This study concludes that signal pattern recognition had potential applications in automated job analysis. The current implementation could be useful for indicating approximate breakpoints, and interactively fine-tuning the breakpoint selection as a means for reducing the time required to perform an analysis. Relevance to industry Job analysis involving large quantities of upper limb biomechanical data can be tedious and prohibitively time-consuming. Automated computerized methods can help make these analyses more practical for use in industry. The current algorithm and implementation can be used as an interactive assistive tool to help reduce the time to do an analysis, and may lead to better automated analysis methods.