Validity and reliability of a voice-recognition game analysis system for field sports

The purpose of this study was to assess the ability of observers to use voice-recognition analysis to accurately classify gait transitions and quantify gait durations typical of team games. Inter-rater and intra-rater reliability was also determined. Four males were filmed performing pre-determined gait protocols. each comprising different sequences of walking, jogging, running and sprinting. Two operators independently classified gait transitions and the time spent in each gait was determined by the voice recognition system. All gait modes as measured by trained observers demonstrated statistically significant correlations (p < 0.01) to pre-determined measurement criteria. The mean absolute error for all gait transitions was less than half a second (0.32-0.36 s) with the maximum percentage error being approximately 4% for the walk, jog and run gaits and 10% for sprinting. Gait classification error was low at 1.9%. The intra-rater and inter-rater reliability was consistently high ranging from r = 0.87 to 0.99. In conclusion, observers using voicerecognition software provided valid measures of time spent in each of the four gait categories with 90% or better accuracy achieved.