Multi-Person Sleeping Respiration Monitoring with
COTS WiFi Devices
(Invited Paper)
Yanni Yang
∗
, Jiannong Cao
∗
, Xuefeng Liu
†
, Kai Xing
‡
∗
The Hong Kong Polytechnic University, Hong Kong, China.
Email: csynyang@comp.polyu.edu.hk, csjcao@comp.polyu.edu.hk
†
Beihang University, Beijing, China. Email: csxfliu@gmail.com
‡
University of Science and Technology of China, Hefei, China. Email: kxing@ustc.edu.cn
Abstract—Recently, non-intrusive respiration monitoring has
attracted much attention. Many respiration monitoring systems
using the commercial off-the-shelf WiFi devices have been de-
veloped. However, these systems mainly have difficulties in the
presence of multiple persons. The difficulty generally comes from
the separation of the effects of multiple persons’ respiration on
the received WiFi signals. Another problem is that even though
the separation can be feasible with some complicated algorithms,
it is still impossible to map the multiple identified respiration
states to the corresponding persons. In this paper, we study
the problem of multi-person sleeping respiration monitoring and
try to address the above challenges. Instead of focusing on
developing complicated signal processing algorithms, we take
another approach: via the deployment of WiFi transceivers. The
key insight comes from the WiFi Fresnel zone model, which
indicates that a carefully placed WiFi transceiver may only
be affected by the person in a certain location. Furthermore,
we consider the sleeping movements of people as well as the
sleeping posture change to improve the robustness of the system.
Extensive experiments show that we can successfully estimate
the respiration rate of multiple persons, with the Mean Absolute
Error (MAE) of 0.5 bpm − 1 bpm.
Index Terms—Respiration monitoring; Multi-person scenario;
WiFi Fresnel zone.
I. INTRODUCTION
Recently, non-intrusive respiration monitoring, particularly
those based on the commercial off-the-shelf (COTS) WiFi de-
vices [1]–[3], attracts much attention. These systems leverage
features extracted from the Channel State Information (CSI) of
WiFi signals, and from which to infer important information
like the respiration rate and the presence of apnea [2], [4].
However, most existing work on WiFi-based respiration
monitoring only works for the single-person scenario and have
difficulties in the presence of multiple persons. One major
difficulty of tracking multiple persons’ respiration states comes
from the fact that, the chest movement of multiple persons will
have accumulative effects on the received WiFi signals that
cannot be easily de-coupled. There are many well-designed
and complicated algorithms proposed to separate multiple
persons’ respiration [5]–[7]. However, these methods generally
rely on the assumption that the respiration rates of multiple
persons are different from each other. Most importantly, ex-
isting work for multi-person respiration monitoring is unable
to map the identified respiration states to the corresponding
persons, which is however of vital importance for performing
targeted health analysis for each person exclusively.
In our work, the above problem is addressed through a new
perspective: via the deployment of WiFi transceivers. The idea
comes from the Fresnel zone model. In [8], [9], the Fresnel
zone model is proposed with regards to respiration monitoring.
According to the Fresnel zone model, a person at different
locations can have different levels of amplitude change for the
respiration pattern on the received WiFi signal. Therefore, to
capture clear and obvious respiration pattern, the deployment
of WiFi transceivers needs to be finely tuned so that the person
is located at the good location in the Fresnel zones.
The basic idea of our approach comes from the observa-
tion that, under certain deployment of WiFi transceivers, the
received WiFi signals can show notable respiration pattern at
some locations in the Fresnel zones. While, at some other lo-
cations, the respiration pattern can be quite obscure. Delighted
by this, we can optimize the deployment of WiFi transceivers,
so that each person is only at the good location of a specific
transceiver pair, meanwhile at the bad location of all the other
transceiver pairs. In this way, we can assign the multi-person
respiration monitoring task to multiple transceiver pairs and
map the identified respiration state of each transceiver pair to
the corresponding person.
However, there are many challenges when realizing the
above idea in practice. First, the deployment of WiFi
transceivers may lose its effectiveness when people move
around and change posture while sleeping, as the respira-
tion pattern is sensitive to the person’s location. To ensure
decent performance, the deployment of transceiver antennas
should still work properly when the person moves. Second,
although the effects of other persons’ respiration on the target
person’s received signals can be decreased to the least under
the optimized deployment of transceivers, it still cannot be
removed completely. However, the minor effects of other
persons’ respiration can lead to the misinterpretation on the
apnea detection for the target person who stops breathing.
To make the deployment more robust when the person
moves, the movement of the person on the bed during sleeping
is configured to comply with Gaussian distribution. Mean-