Making waves with WiFi data

“My dataset took a lot of time and effort to col­lect and process. I make my data avail­able online in the hope that it pre­vents oth­ers from wast­ing time rein­vent­ing the wheel and dupli­cat­ing my efforts,” says Uni­ver­si­ty of Twente researcher, Jeroen Klein Brinke.

Jeroen Klein Brinke is a third year PhD stu­dent work­ing in the Per­va­sive Sys­tems Group at the Uni­ver­si­ty of Twente.

After study­ing his Bachelor’s and Master’s degree in Tech­ni­cal Com­put­er Sci­ence, his doc­tor­al research inves­ti­gates device-free (wire­less) sens­ing tech­nolo­gies and specif­i­cal­ly how human activ­i­ty affects the behav­iour of radio waves. 

“Many homes and build­ings have WiFi net­works that can be used to detect and iden­ti­fy var­i­ous human activ­i­ties,” says Jeroen.

“The prop­a­ga­tion of radio waves in space can pro­vide infor­ma­tion about the tasks and activ­i­ties peo­ple car­ry out in an envi­ron­ment.”

Wireless sensing for the societal benefit

Jeroen explains that such wire­less sens­ing tech­nolo­gies can be used to pro­vide health­care for the elder­ly. 

“Elder­ly peo­ple who live alone at home are at risk of poor phys­i­cal and psy­cho­log­i­cal health. They often expe­ri­ence dif­fi­cul­ties per­form­ing dai­ly tasks in social iso­la­tion and are at increased risk of falling and injur­ing them­selves,” says Jeroen.

He adds that this places a high bur­den on care­givers.

“An elder­ly per­son may require fre­quent vis­its from car­ers or con­stant sur­veil­lance in order to pre­vent a fall or injury. Wire­less sens­ing tech­nolo­gies can­not replace human care entire­ly, but it can be used to inform car­ers about a person’s mobil­i­ty and whether they are like­ly to fall so that pre­ven­ta­tive mea­sures can be tak­en in advance.”

Jeroen’s top downloaded dataset

Jeroen’s dataset, ‘Chan­nel state infor­ma­tion (WiFi traces) for 6 activ­i­ties’, was cre­at­ed dur­ing his Master’s degree research project. Since it was pub­lished in 4TU.ResearchData, it has been down­loaded more than 900 times and was one of the top down­loaded datasets in March… [Con­grat­u­la­tions, Jeroen!]

“I’m an advo­cate for data reuse,” says Jeroen. “If I need a spe­cif­ic dataset, the first thing I always do is check online data repos­i­to­ries to see if a suit­able dataset is avail­able.”

He con­tin­ues, “When I start­ed my research I found that exist­ing datasets includ­ed WiFi traces from a sin­gle par­tic­i­pant per­form­ing one activ­i­ty for a day. These proof-of-con­cept datasets sim­ply demon­strate that wire­less sens­ing tech­nol­o­gy works, but I want­ed to build upon these stud­ies and advance the research field by cre­at­ing a more com­plex and var­ied dataset that’s accom­pa­nied by rich doc­u­men­ta­tion so that oth­ers can reuse and repro­duce it.”

Jeroen’s dataset con­tains bina­ry and MATLAB files of radio waves for nine par­tic­i­pants in a stu­dent room over six days con­duct­ing six activ­i­ties; sit­ting, clap­ping, wav­ing, jump­ing, falling and walk­ing. The radio waves were col­lect­ed using a net­work node com­pris­ing a sin­gle trans­mit­ter and receiv­er pair using mul­ti­ple anten­nas. 

Jeroen devel­oped a deep learn­ing algo­rithm to process the WiFi traces and dis­crim­i­nate between the dif­fer­ent activ­i­ties. His pub­lished dataset also con­tains a tool to allow oth­er researchers to visu­alise the data more eas­i­ly. 

Aside from want­i­ng to make his orig­i­nal con­tri­bu­tion to his research field, Jeroen was moti­vat­ed to pub­lish his data because he under­stands the chal­lenges of col­lect­ing data from human par­tic­i­pants. 2802

“My dataset took a lot of time and effort to col­lect and process. Uphold­ing eth­i­cal con­sid­er­a­tions asso­ci­at­ed with the col­lec­tion of per­son­al data, such as eth­i­cal approval, informed con­sent, con­fi­den­tial­i­ty and anonymi­ty isn’t a triv­ial respon­si­bil­i­ty. By mak­ing my data avail­able for reuse, I hope that I can pre­vent oth­ers from wast­ing time rein­vent­ing the wheel and dupli­cat­ing my efforts.”

Jeroen was proud to receive emails about his dataset from researchers around the world soon after its pub­li­ca­tion online. “Researchers work­ing on sim­i­lar projects con­tact­ed me to request more infor­ma­tion about how I col­lect­ed the data dur­ing my exper­i­ments and asked for point­ers on how they could inter­ro­gate the dataset which was ful­fill­ing and reward­ing. It just shows how valu­able a rel­a­tive­ly small dataset can be!”

The future of wireless sensing 

The next steps in Jeroen’s research will inves­ti­gate how to upscale wire­less sens­ing tech­nolo­gies at low cost for a vari­ety of real-world appli­ca­tions.

“In addi­tion to health care, wire­less sens­ing can be used to mon­i­tor ani­mals in pro­duc­tion sys­tems, such as baby piglets, or to mon­i­tor insect bio­di­ver­si­ty, for exam­ple. Such appli­ca­tions require tai­lored approach­es. For instance, due to the size dif­fer­ence, the activ­i­ty of piglets could be mon­i­tored using radio waves, where­as insects would require high fre­quen­cy mil­lime­tre waves,” he explains. 

As the pop­u­lar­i­ty of wire­less sens­ing tech­nolo­gies grows in the future, researchers must be mind­ful about the risk of its abuse. 

“Wire­less sens­ing is extreme­ly pow­er­ful,” says Jeroen. “WiFi traces are not nec­es­sar­i­ly localised to one build­ing but can prop­a­gate walls. They can also detect breath­ing and heart rate, so, there’s an oppor­tu­ni­ty to col­lect high­ly sen­si­tive per­son­al data using this tech­nol­o­gy. It’s, there­fore, imper­a­tive that researchers uphold gen­er­al data pro­tec­tion reg­u­la­tions and act respon­si­bly when man­ag­ing this data.”

Jeroen states that he will con­tin­ue to pub­lish his research data in the 4TU.ResearchData repos­i­to­ry and is in con­tact with Mar­i­an­na Avetisyan, data stew­ard for the fac­ul­ty of Elec­tri­cal Engi­neer­ing Math­e­mat­ics and Com­put­er Sci­ence, for advice on how to pre­pare his future per­son­al WiFi trace data for pub­li­ca­tion online. 

Thank you, Jeroen, for shar­ing your research sto­ry with us and con­grat­u­la­tions on your top down­loaded dataset!

Writ­ten by Con­nie Clare (4TU.ResearchData)