Table 1.

Summary of the main validation studies analyzing motion capture accuracy of the Kinect (v1 and v2) and of wearable IMUs

Refs.	Sensor(s)	Participants	Exercises	Parameters	Applications	Performance
[49]	Kinect v1	1H	6 primary movements	Joints positions, bone lengths	At-home elderly rehabilitation	RMSE $<10cm$
[50]	Kinect v1	20H	Reaching, standing balance	Joints positions	Postural control	ICC $<0.16$ CV $<11.6\%$
[51]	Kinect v1	48H, 2 sessions	4 primary movements	Joints angles	Funct. assessment	RMSE $<9\%$
[52]	Kinect v1	10H + 9PD	12 primary and functional movements	Joints positions and angles, timings	PD assessment	LoA $<10\%$ ICC $>0.9$
[53]	Kinect v1	20H	Standing sway	Marker positions	Exergaming	SDev $<30\%$
[54]	Kinect v1	15H	Squat	Lower limb joints angles	Kinect evaluation	RMSE $<5^{\circ }$ ICC $>0.9$
[36]	Kinect v2	30H, 2 sessions	Standing balance	Trunk and pelvis angles	Balance and postural control	r $>0.75$ ICC $>0.7$
[42]	Kinect v2	30H, 2 sessions	Walking	Lower limbs joints angles	Gait analysis	$0.4<$ r $<0.75$
[39]	Kinect v2	20H	5 primary shoulder movements	Upper limbs joints angles	Clinic and home rehabilitation	RMSE = $3.9^{\circ }$
[43]	Kinect v2	20H	Walking	Stance and step time, step length, time, velocity	Gait analysis	LoA $<11\%$ ICC $>0.9$
[41]	Kinect v2	30H	Standing and sitting exercises	Upper body joint positions	Rehabilitation	RMSE $<5\%$ r $>0.9$
[55]	4 IMU (custom) on thighs, shanks	3H	Walking	Hip and knee joint angles	Gait analysis	RMSE $<8.7^{\circ }$ r $>0.7$
[6]	2 IMU (Xsens MTx) on pelvis, thigh	20H	Walking	Hip joint angle	Gait analysis	LoA $=1.6^{\circ }$
[56]	8 IMU (Xsens MTx) on thorax, lower back, thighs, shanks, feet	9H (children)	Static postures	Hip and knee joint angles	Knee amputees and cerebral palsy	RMSE $=1.8^{\circ }$
[57]	6 IMU (not specified) on pelvis, thigh, shank	9H + 1INJ	Running and outdoor training routines	Activity classification, hip and knee joints angles	Athlete training, injury prevention	98%
[58]	3 IMU (Shimmer) on thigh, shank, foot	58REH	Primary movements	Exercise recognition	Clinic and home rehabilitation	83%
[8]	5 IMU (Riablo) on trunk, thigh, shank	22H, 2 sessions	4 knee exercises	Hip and knee joint angles	Clinic and home rehabilitation	RMSE $=3.1^{\circ }$
[59]	2 IMU (Opal) on lower back, upper arm	6H	ADL routines, walking	Trunk and upper limb joints angles	IMU evaluation	RMSE $<5.5^{\circ }$
[60]	6 IMU (Xsens MTx) on thigh, shank, foot	1AMP	Walking	Knee angle	Gait analysis	RMSE $<3^{\circ }$
[61]	7 IMU (Opal) on trunk, upper arm, forearm, hand	8H	2 swimming styles	Upper limb joints angles	swimming analysis	RMSE $<7^{\circ }$
[62]	3 IMU (Opal) on waist, thigh, shank, 1 resistive strip	14H	Sit-to-stand, walking	Exercise timings, stride length	Clinic and home rehabilitation	r $>0.7$ ICC $>0.95$
[63]	17 IMU (Synertial) full body	20H	5m timed up and go	Full body joints	Clinic evaluation	RMSE $<13.6^{\circ }$
[64]	2 IMU (APDM) thigh and shank	18REH	Knee ROM, walking	Knee angle	Clinic rehabilitation	–
[9]	17 IMU (Xsens MVN) full body	12H	Functional movements, work actions	Full body model	Occupational biomechanics	RMSE $<5^{\circ }$
[65]	3 IMU (I2M MT) on pelvis, trunk, upper arm	6H lab + 10H workplace	Work actions	Trunk orientation, shoulder joint angles	Occupational biomechanics	RMSE $<12.1^{\circ }$
[47]	Kinect v1, Kinect v2	20H (v1) + 20H (v2)	15 static poses	Joints positions	Posture evaluation	RMSE(v1) $=76mm$ RMSE(v2) $=87mm$
[48]	Kinect v1, Kinect v2	10H, 3 camera angles	Sitting and standing postures	Joints positions	Motion capture	RMSE $<100mm$
[33]	Kinect v1, Kinect v2	13H	ROM and static stretching exercises	Joints positions	Rehabilitation	RMSE $=15mm$(wrist) RMSE $<10mm$(other)
[66]	Kinect v1 + 2IMU (InertiaCube3) on upper limb	1H	Hand-to-mouth	Upper limb position	Motion capture	–
[67]	Kinect v1 + 1 x-IMU	1H	Knee flexion/extension	Knee angle	Motion capture	RMSE $<14.2^{\circ }$
[68]	Kinect v1 Vs Kinovea (MBS) Vs Hillcrest (IMU)	2H	Walking	Hip and knee angles	Gait rehabilitation	RMSE $=3^{\circ }$ (IMU) RMSE $={10}^{\circ }$ (Kin.)

Participants are labeled as healthy (H), Parkinsons (PD), injured (INJ), in rehabilitation (REH) and amputees (AMP). All experiments were conducted in lab unless otherwise stated. Reported performance metrics include Root Mean Square Error (RMSE), Interclass Correlation Coefficient (ICC), Coefficient of Variation (CV), Limits of Agreement (LoA), Standard Deviation (SDev) and Pearson’s correlation coefficient (r)