Pos estimation using ESP32

This program is build for estimate Position using ESP32.

icanfi

---

Under module icanfi, it cantain the function to load csv file of capture CSI data, preprocess it and some filter utility.

load.py, new_load.py and multi_load.py

To load csv file that is capture from idf monitor you can use icanfi.load to do the trick. If you use the utility provide in connection -> serial_load.py please rember to use icanfi.new_load to help you througt the work.

utility	description
icanfi.load(filename,mod = 'a')	use it to load idf monitor file mod stands for capture data type: 'a' amplitude, 'p' phase
icanfi.new_load(filename,mod = 'a')	use it to load serial_load.py file mod stands for capture data type: 'a' amplitude, 'p' phase
icanfi.multi_load(filename,mod = 'a')	use it to load the data that is collected from raspberry pi, it can help you organize links in order. mod stands for capture data type: 'a' amplitude, 'p' phase

After you load the csv file, the data will be store in format below:

#for load and new_load
#dict(np.array)

CSI_data = {
    "time" : np.array([*time sequence*]),
    0 : np.array([*CSI data sequence*]), #subcarrier index - 1
    1 : np.array([*CSI data sequence*]),
    2 : np.array([*CSI data sequence*]),
    ...
    63 : np.array([*CSI data sequence*])
}

#for multi_load
#dict(dict(np.array))
link = {
    1  :  {
            "time" : np.array([*time sequence*]),
            0 : np.array([*CSI data sequence*]), #subcarrier index - 1
            1 : np.array([*CSI data sequence*]),
            2 : np.array([*CSI data sequence*]),
            ...
            63 : np.array([*CSI data sequence*])
        },

    2  :  {
            "time" : np.array([*time sequence*]),
            0 : np.array([*CSI data sequence*]), #subcarrier index - 1
            1 : np.array([*CSI data sequence*]),
            2 : np.array([*CSI data sequence*]),
            ...
            63 : np.array([*CSI data sequence*])
        },

    3  :  {
            "time" : np.array([*time sequence*]),
            0 : np.array([*CSI data sequence*]), #subcarrier index - 1
            1 : np.array([*CSI data sequence*]),
            2 : np.array([*CSI data sequence*]),
            ...
            63 : np.array([*CSI data sequence*])
        },

    4  :  {
            "time" : np.array([*time sequence*]),
            0 : np.array([*CSI data sequence*]), #subcarrier index - 1
            1 : np.array([*CSI data sequence*]),
            2 : np.array([*CSI data sequence*]),
            ...
            63 : np.array([*CSI data sequence*])
        }
}

The idea of link index

In this setup the link index will be

device	index
RASP0 esp 0	1
RASP0 esp 1	2
RASP1 esp 0	3
RASP1 esp 1	4

                    RASP 0
                   /    \
                  /      \
                esp 0    esp 1

        esp 0                   AP
         /                       \
        /                         \
    RASP 1                         power
        \                         /
         \                       /
        esp 1                   AP

                AP       AP
                 \       /
                  \     /
                   power

analysis.py

After load in the data set, you can use utilities in analysis.py to check the data. (note: It only provides minimum support. If there is an error, the quickest way to fix it is to write you own analysis system using matplotlib.pyplot)

utility	description
icanfi.analysis_time(data,Lsub = 0,Usub = SUBCARRIER,Sframe = 0,Eframe = None)	analysis_time can help us to understand how subcarrier behave in time domain. Lsub: lower subcarrier, Usub: upper subcarrier, Sframe: starting frame, Eframe: ending frame
icanfi.analysis_sub(data,Lsub = 0,Usub = SUBCARRIER,Sframe = 0,Eframe = None)	analysis_sub demonstrate the data by subcarrier, by this we can observie the amplitude difference between each subcarrier in each frame. Lsub: lower subcarrier, Usub: upper subcarrier, Sframe: starting frame, Eframe: ending frame

preprocess.py

After load in the data set, preprocess can help us to transfrom the dataset into desire form. you can use icanfi.downsampling to get uniform data distribution, icanfi.remove_DC to eliminate DC component... the uilities is explain below.

utility	description
icanfi.into_energy(data)	Transform data into energy of the signal in channel, please do this opperation in your LAST step of preprocessing, otherwise you might have trobule in doing further work. return dict(np.array()) #check source code or test.py to get better undestanding
icanfi.delta_R(data_energy)	compute the difference between frame, take input from into_energy. return dict(np.array)
icanfi.windowing(data,windowsize = WINDOWSIZE,hop = HOP,drop = DROP)	Do windowing operation. windowsize: how many second can be contain in a window defult is 20 second hop: the starting second from the previous one defult is 3 second. drop: if the gap between frame is larger them drop the window is discard return list(dict(np.array()))
icanfi.downsampling(data,Srate = DOWNSAMPLING_S,bias = BIAS,extand = False)	downsapling will modify the original data,i.e. no need to assign another parameter Srate: sample rate bias: tolerance for data bias, the data have to be in range of second +/- bais. extend: acceptance of extending the data if there is a miss *interact with windowing can help you filter out those window that has large miss

filters.py

utilities in filters.py are design for data that process by into_energy,so be ware of the format.

utility	description
icanfi.lowpass(data)	to change low pass filter's cut of frequency, please change it in parameter.py
icanfi.bandpass(data)	to change band pass filter's cut of frequency, please change it in parameter.py
icanfi.data_fft(data)	do fourier transform and return result
icanfi.wavelet	do wavelet transform return the result

algorithm.py

contain some PCA utility, which is unrelated to this project ,but still helpful for building a new PCA utility.

parameter.py

all parameter can be set up here.

#our CSI model can only detect 20HZ max because the serial speed
#butterworth filter


BUTTER_N = 7
BUTTER_WN = 0.5 #cut off frequency
BUTTER_OUTPUT = 'sos'
BUTTER_ANALOG = False
BANDPASS_WN = [0.05,1]
#downsampling rate
DOWNSAMPLING_S = 0.04
DOWNSAMPLING_RATE = 1/DOWNSAMPLING_S
BIAS = 0.01

#list the subcarriers that has no informationn about respiratory
NO_USE_SUB = [0 ,1 ,2 ,3 ,4 ,5 ,32 ,59 ,60 ,61 ,62 ,63 ,64 ,65 ,123 ,124 ,125 ,126 ,127 ,128 ,129 ,130 ,131 ,132 ,133 ,191]

#threshold for SNR subcarrier that is been selected
SNR_SUB_THRESHOLD = 5000000
CHOOSE_SUB = 5
THRESHOLD_DECLINE = 0.95


#the amount of subcarrier in one channel

SUBCARRIER = 64

#wavelet transform

LEVEL = 6
WAVELET = 'db6'

#windowing
WINDOWSIZE = 20
HOP = 3
DROP = 100

connection system (new version is in realtimeconnection)

---

To have better control over whole data collecting system, I use lan to connect the CSI capturing device. the system is being saparated into two programs, connection and ras_connection

connection

connection can be use in Windows os, control.py is all you need for you to connect to raspberry pi. two other can help you test ESP32 if there is something unexpected.

ras_connection

ras_connection is designed for rapberry pi,rasp.py can help you get the ESP32 data.

system overview

        [Control Center]                  [Data Collector]
                |                                 |
                |---------check connection------>>|
                |                                 |
                |<<--------send connected---------|
                |                                 |
          [check reply]                           |
                |                                 |
          [setfile name]                          |
                |                                 |
                |-----------send start---------->>|
                |                                 |
                |                     [start collecting process]
                |                                 |
                |<<-----send collected data-------|
                |                                 |
        [save as csv file]                        |
                |                                 |
                |-----------send stop----------->>|
                |                                 |

To run the collection system, you should run the rasp.py inside the ras_connection first. Then make sure all you collecting device is up. After that you should run control.py to initiate the data collecting process.

For more detail or small change in device set up. check the source code, you can change the parameter to meet up your needs.