Table 1.

Abbreviations and symbols.

Abbreviations and symbols	Description	Abbreviations and symbols	Description
3D	Three dimensions	ANN	Artificial neural network
ABW	Activity-based windowing	BSS	Blind source separation
ADL	Activities of daily living	CRF	Conditional random field
AFE	Analogue front end	DLC	Deep learning-based classification
CCA	Canonical correlation analysis	DLS	Deep learning-based semisupervised model
CFS		DLF	Depp learning-based features
CNN	Convolutional neural network	DBN	Dynamic Bayesian network
CPD	Point change detection	EM	Expectation-maximization
CSS	Contact switch sensors	FA	Factor analysis
DBN	Deep belief network	FP	False positives
DFT	Discrete Fourier transform	FN	The number of false negatives
DL	Deep learning	GMM	Gaussian mixture model
DT	Decision tree	ICA	Independent component analysis
HAR	Human activity recognition	LS	Least squares
HARS	Human activity recognition system	NB	Naïve Bayes
HMM	Hidden Markov model	RF	Random forest
IMU	Gyroscope, accelerometers, and magnetic sensors	RBF	Time complexity in modeling
KNN	K-nearest neighbor	RBM	Restricted Boltzmann machine
LDA	Linear discriminant analysis	SBHAR	Smartphone-based HAR
L-SSW	Last-state sensor windowing	TCM	Time complexity in modeling
LSTM	Long short-term memory	Radial basis function	TCR time complexity in recognition
MEMS	Microelectromechanical systems	w i	The ratio of class i in all samples
Mhealth	Mobile health	F	Freight gate
NN	Neural network	i t , ot and ft	Input, output, and forget gates considered in time t, respectively
PCA	Principal component analysis	h (all)	Hidden values
PI	Passive infrared	Recalli	Sample ratio of class i that is correctly predicted on all correct samples
PN	Number of participants	K	Kernel function
PWM	Pulse width modulation	N	The total number of all samples
QDA	Quadratic discriminant analysis	Precisioni	The ratio of an instance of class i that is correctly predicted on all predicted samples
REALDISP	REAListic sensor DISPlacement	b i , bf, bc and bo	Bias vectors
RFID	Radio frequency identification	c t−1	Cell output at the previous time stage
RNN	Recurrent neural network	W ai , Whi, Wci, Waf, Whf, Wcf, Whi is hidden-input gate matrix Wac, Whc, Wao, Who, Wco	Matrixes of weight: Wai is input-input gate matrix, Whi is hidden-input gate matrix, and the rest of the W is named in this way
STEW	Sensor dependency extension windowing	c t	The state of memory at time t
SDW	Sensor-dependent windowing	O	Output gate
SEW	Sensor event-based windowing	I	Input gate
SHCS	Smart healthcare system	C	Cell activation vectors
SVM	Support vector machine	n i	The number of samples in ith class
TBW	Time-based windowing	a t	Input to the memory cell layer at time t
TP	The number of true positives	All σ	Non-linear functions
TSW	Time slice-based windowing