Fall detection method and system

1. A fall detection method, comprising:

receiving, by a first receiving antenna, a first WiFi signal stream propagating through an environment;

receiving, by a second receiving antenna, a second WiFi signal stream propagating through the environment;

determining a physical layer Channel State Information (CSI) stream, namely, a first CSI stream, of the first WiFi signal stream;

determining a physical layer Channel State Information stream, namely, a second CSI stream, of the second WiFi signal stream;

determining a phase difference, namely, a CSI phase difference, between a respective phase of the first CSI stream of the first WiFi signal stream and the second CSI stream of the second WiFi signal stream at the same time point, to form a CSI phase difference stream; and

determining, according to the CSI phase difference stream, a fall event.

2. The fall detection method according to claim 1 , further comprising determining, according to the first and second CSI streams and the CSI phase difference stream, the fall event.

3. The fall detection method according to claim 2 , further comprising determining if a raw CSI phase difference signal and a filtered CSI phase difference signal are in a fluctuation state or a stable state by using a threshold-based sliding window method; and

for the raw CSI phase difference signal and the filtered CSI phase difference signal, detecting the transition from the fluctuation state to the stable state, and determining a finishing reference point of fall and fall-like activities by checking if the two signals enter the stable state and based on a time difference between the raw CSI phase difference signal and the filtered CSI phase difference signal when entering the stable state.

4. The fall detection method according to claim 3 , further comprising:

extracting the following feature of the raw CSI phase difference stream and/or the filtered CSI phase difference stream: power decline ratio (PDR), which is determined to be the energy decline ratio within a predetermined frequency range before and after a predetermined time length of a base time point over a time-frequency spectrum, respectively, wherein the base time point is the determined finishing reference point of fall and fall-like activities; and

determining, according to the raw CSI phase difference stream and/or the filtered CSI phase difference stream, the fall event.

5. The fall detection method according to claim 4 , wherein further comprising:

extracting the following features of the first CSI stream or the second CSI stream:

a normalized standard deviation (STD), a median absolute deviation (MAD, an offset of signal strength, interquartile range (IR), signal entropy, and a velocity of signal change;

extracting the following features of the raw CSI phase difference stream and/or the filtered CSI phase difference stream: the normalized standard deviation (STD), the median absolute deviation (MAD), the offset of signal strength, interquartile range (IR), signal entropy, and the velocity of signal change; and

determining, according to the features extracted by a feature extraction device, by using a one-class Support Vector Machine (SVM), the fall event.

6. The fall detection method according to claim 1 , further comprising:

identifying a finishing reference point of fall and fall-like activities according to the CSI phase difference stream, and determining a starting reference point of fall and fall-like activities according to a trace back window size.

7. The fall detection method according to claim 3 , wherein if and only if the raw CSI phase difference signal and the filtered CSI phase difference signal enter the stable state with the time difference, namely, a time lag, that is less than a predetermined threshold, determining the finishing reference point of fall and fall-like activities.