Wi-Motion: A Robust Human Activity Recognition Using WiFi Signals

Abstract

Recent research has shown that human motions and positions can be recognized through WiFi signals. The key intuition is that different motions and positions introduce different multi-path distortions in WiFi signals and generate different patterns in the time-series of channel state information (CSI). In this paper, we propose Wi-Motion, a WiFi-based human activities recognition system. Unlike existing systems, Wi-Motion adopts the amplitude and phase information extracted from the CSI sequence to construct the classifiers respectively, and combines the results using a combination strategy based on posterior probability. As the simulation results shows, Wi-Motion can recognize six human activities with the mean accuracy of 98.4%.

Human Activity Sensing with Wireless Signals: A Survey

“…Table 4 provides a summary of the models between human activity characteristics and critical parameters of wireless signal propagation. [36,69], Coarse-grained estimation [14,18,19,[23][24][25][26]37,40,42,43,46,47,49,50,[52][53][54] Distance (dLoS) Fresnel zone model [106][107][108][109][110] Direction Fresnel zone model [106], AoA with antenna array [11,15,65,[111][112][113][114][115] aw Signal Distance, Direction, Velocity mD-Track [116] 3.1. Phase…”

“…PCA-based subcarrier selection: The CSI measurements are highly correlated among subcarriers, and different subcarriers have different sensitivity for a given activity [42]. Thus, the most common method for redundant removal is to apply principal component analysis (PCA) on the CSI measurements to extract principal components.…”

“…Existing researches hold different views on principal component (PC) selection. Some solutions select the first PC, which contains the highest eigenvalue among all the PCs and may correspond to the features caused by human motions [42,58]. On the contrary, some researches choose to discard the first PC and preserve the second PC [36,80].…”

“…Time-domain Pros: relatively simple calculation; Cons: vulnerable to environmental changes and noise maximum, minimum, mean, standard deviation, kurtosis, skewness, variance, median and median absolute deviation, percentiles, root sum square, interquartile range [13,18,37,40,42,43,46,[48][49][50]62,72,73,75,76,[78][79][80][83][84][85][88][89][90]97,108,127,130], time lag [37], power decline ratio [37], amplitude sequence [20,54,74,91], phase sequence [39,51,107], CV [38] Frequency-domain Pros: capture the periodical characteristics of human motion; Cons: large amount of calculation FFT coefficient [12,83], dominant frequency [43], power spectrum density [88], spectral entropy [43,57,70,72,…”

“…The statistics feature often characterize the shape of the received waveform in the time domain. There are a large amount of studies [13,37,40,42,43,70,72,73,76,79,80,83,85,[88][89][90]108] that extract time-domain statistical features (maximum, minimum, mean, standard deviation, kurtosis, skewness, variance, median and median absolute deviation, percentiles, root sum square, interquartile range) from amplitude or phase streams. In particular, WIG [78] extracts the above time-domain features from anomaly data series obtained by the LOF-based segmentation method.…”