Table 2.

Comparison of various wearable FoG detection systems in the literature

	Sensor				Algorithm		Data		Performance (%)			Split
Work (ref. no.)	Type	Location	Cues	CP	ML	Nf	Subjects	FoG	Sens	Spec	Accu
Bächlin et al., 2010	Acc	Multi	A	×	×	2	10	237	73.1	81.6	—	×
Mazilu et al., 2012	Acc	Multi	A, S	√	√	7	10	237	62.1	95.2	—	√
Pepa et al., 2015	Acc	Waist	A	√	×	3	18	73	82.3	76.8	—	×
Kim et al., 2015	Acc	Multi	A, S	√	√	12	9	—	86.0	91.7	—	√
Lorenzi et al., 2016	IMUs	Shin	A	×	×	1	16	—	94.5	96.7	95.6	×
Ahn et al., 2017	IMUs	Glass	V	×	×	1	10	42	97.0	88.0	—	×
C. Punin et al., 2017	IMUs	Ankle	S	√	×	2	8	27	86.7	60.6	—	×
(Kita et al., 2017)	IMUs	Shin	A, S	√	×	1	32	—	97.6	93.4	—	×
Mikos et al., 2019	IMUs	Shin	A, S	×	√	3	63	485	95.6	90.2	—	√
This work	PSUs	Insoles	A, V	√	×	2	30	1,502	97.6	95.0	95.2	√

Abbreviations: Sensor, accelerometers (Acc), inertial measurement units (IMUs), pressure sensing units (PSUs). Sensor Location, the position of sensors, Multi means multiple sensors located chest, waist, knee, ankle, and so on. Cues, auditory (A), somatosensory (S), visual (V). CP, cellphone used to collect data and run algorithms. Algorithm, machine learning (ML), number of features employed (Nf). Data, total number of subjects (Subjects), the total number of FoG occurrence (FoG). Performance, sensitivity (Sens), specificity (Spec), Accuracy (Accu). The value of accuracy lies between the values of sensitivity and specificity. Split, Split the dataset into a training and testing set.