Time-Frequency Linearization of Reactive Cortical Responses for the Early Detection of Balance Losses

Table 1

Overview PIFD strategies at the state of the art.

Ref.	Technology	Fall indicators	Class. Algorithm	Fall type	Performance
[13]	Trunk-positioned 3D ACC	Trunk vertical velocity (TVV)	Single Thr:	n. 3 SF: FwF, BwF, SdF
[25]	Waist-positioned 3D GYR and ACC	Waist vertical velocity (WVV)	Single Thr: WVV in nonfall activities	n. 4 SF: FwF, BwF, SdF, OrF
[26]	9DoF IMU by Xsens Tech. positioned on chest	Acceleration and angular velocity of the chest segment	Multiple Thr: (1); (2)	n. 3 SF: FwF, BwF, SdF
[14]	Trunk-positioned 3D ACC	Acceleration Time Series (ATS) of the chest	ML: (1) The extracted ATS trains a HMM; (2) The HMM outcome is compared with a single Thr	n. 2 SF: FwF, SdF
[23]	Waist-positioned 3D GYR and ACC	Acceleration and angular velocity of the waist segment	ML: (1) Mean and variance of the axis acceleration and angular velocity; (2) Classification by SVM	n. 4 SF: FwF, BwF, SdF, falling from sit; IF: slip-induced fall
[17]	EEG wireless headset+surface EMG	EEG Power Spectrum Density (PSD) level in BP, , rhythms	Multiple Thr: (1) The EMG is used as a trigger for cortical analysis; (2) The EEG PSD is evaluated in BP, , bands; (3) The levels are compared with history-based thresholds	n. 1 SF: BwF; IF: loss of balance on weighbridge	(from gastrocnemius contraction)
[9]	MCS: total body monitoring	Acceleration and vertical velocity of upper arms, trunk, tibia, and head	Multiple Thr+statistical model: threshold based on a ARIMA model based on data history	IF: slip-induced fall
[10]	MCS: total body monitoring	Acceleration of all the monitored body segments	ML: (1) The accelerations are analyzed by ICA; (2) A neural network is used to distinguish walk from perturbations	IF: slip-induced fall
[11]	MCS+trunk-positioned IMU sensor	Sagittal angle and angular velocity of the trunk	Multiple Thr+statistical model: threshold based on a AR model based on data history	IF: slip-induced fall
[15]	Hip encoder on active pelvis orthosis	Hip angle	Single Thr: increment of the error function between the current hip angles (from encoder) and the ones provided by a pool of adaptive oscillators	IF: slip-induced fall

Technology acronyms—ACC: accelerometer; GYR: gyroscope; MCS: motion capture system; IMU: inertial measurement unit; DoF: degree of freedom. Class. algorithm acronyms—Thr: threshold; ML: machine learning; HMM: Hidden Markov Model; SVM: support vector machine; BP: Bereishaft potential (EEG); ARIMA: autoregressive integrated moving average; ICA: independent component analysis; AR: autoregression. Type of fall acronyms—SF: simulated fall; IF: involuntary fall (unexpected); FwF: forward fall; BwF: backward fall; SdF: lateral fall; OrF: fall from orthostatic position.