Table 4.

RMSE, nRMSE, and r (mean ± standard deviation) between model predictions and ground truth kinematics for models trained on a single activity and tested on that same activity. Bold indicates most accurate model architecture for that activity.

Metric	Model	Train: LW Test: LW	Train: RA Test: RA	Train: RD Test: RD	Train: SA Test: SA	Train: SD Test: SD
RMSE°	MLR	9.6 ± 3.5°	31.2 ± 10.6°	13.8 ± 2.4°	7.9 ± 3.5°	7.9 ± 1.3°
	CNNLSTM	6.2 ± 2.2°	10.3 ± 4.5°	8.3 ± 1.4°	13.4 ± 5.2°	18.8 ± 6.8°
	BiLSTM	5.5 ± 1.6°	8.2 ± 2.9°	7.0 ± 2.0°	5.3 ± 1.7°	7.2 ± 2.1°
	BioMAT	5.3 ± 2.1°	6.7 ± 2.0°	6.9 ± 2.2°	4.9 ± 1.4°	5.6 ± 1.3°
nRMSE	MLR	13.1 ± 3.2	33.2 ± 6.7	16.2 ± 2.2	9.5 ± 2.7	10.1 ± 3.6
	CNNLSTM	8.4 ± 1.6	10.6 ± 2.3	9.7 ± 1.2	16.0 ± 1.5	22.4 ± 0.9
	BiLSTM	7.5 ± 0.3	8.6 ± 1.2	8.2 ± 2.4	6.5 ± 1.2	9.5 ± 5.1
	BioMAT	7.3 ± 2.3	7.1 ± 0.6	8.2 ± 3.1	5.9 ± 0.7	7.5 ± 3.9
r	MLR	0.91 ± 0.04	0.90 ± 0.03	0.83 ± 0.06	0.96 ± 0.04	0.95 ± 0.02
	CNNLSTM	0.94 ± 0.03	0.91 ± 0.04	0.88 ± 0.03	0.70 ± 0.30	−0.02 ± 0.04
	BiLSTM	0.97 ± 0.03	0.96 ± 0.01	0.93 ± 0.03	0.98 ± 0.03	0.98 ± 0.01
	BioMAT	0.97 ± 0.02	0.97 ± 0.01	0.95 ± 0.03	0.97 ± 0.03	0.98 ± 0.02

LW: Level Walking, RA: Ramp Ascent, RD: Ramp Descent, SA: Stair Ascent, SD: Stair Descent.