Research Article

Accelerometer-Based Automated Counting of Ten Exercises without Exercise-Specific Training or Tuning

Table 2

Root mean square error (RMSE) of counts for each analysis technique.

ExerciseThreshold crossingThreshold with low passFourierAvg RMSE

Arm circle1.999.658.087.36
Bicep curl16.707.3612.2912.71
Bridge13.9414.397.0712.27
Crunch7.328.5612.189.58
Elbow extension13.7010.2010.5711.59
Lower trunk rotation13.479.977.4510.59
Lunge16.951.830.769.85
Push-ups10.531.393.706.49
Squats10.711.500.936.27
Upper trunk rotation19.069.7714.1114.81

Avg RMSE13.388.699.00ā€‰

Threshold crossing with a low-pass filter performs best with an average root mean square error of 8.69, with the Fourier transform being close with 9.00.