TABLE XV.

Summary of Activity and Gesture Detectors for Microcontrollers

Method	Sensor(s)	Task	Accuracy	Flash
GesturePod ProtoNN [212]	MPU6050 inside white cane	Detect 5 cane gestures	92%	6 kB
Auritus FastRNN [2]	eSense earable	Detect 9 macro activities	98%	6 kB
Bian et al. 1D-CNN [213]	Wrist-worn capacitive array	Detect 7 hand gestures	96%	30 kB
T’Jonck et al. CNN [214]	BMA400 inside mattress	Detect 5 bed activities	89%	< 1 MB
Zhou et al. HDC + SVM [215]	MPU-6050 and EMG pad (wrist-worn)	Detect 13 hand gestures across 8 limb positions	93%	135 kB
Elsts et al. CNN [216]	Colibri Wireless IMU	Detect 18 macro activities	73% (F1 score)	20 kB
Coelho et al. DT [217]	Chest, waist and ankle-mounted IMU	Detect 12 macro activities	97%	22 kB
FastGRNN (LSQ) [57]	IMU on torso and limbs	Detect 6 macro activities and 19 sports activities	96% 84%	3 kB 3.25 kB