Zhang Daqing's research group in the School of Computer Science has made important progress in the field of ubiquitous wireless perception research

Recently, Professor Zhang Daqing's research group of the School of Computer Science of Peking University has made a series of important progress
in the mapping relationship between the Fresnel zone perception model and Doppler speed, speed and position orientation, the perception boundary of wireless perception, the perception limit, the perception signal quality and the perception model of moving scenes, focusing on the basic theoretical problems in the field of intelligent wireless perception based on WiFi and radar 。 Among them, 6 long articles were published in the IEEE Journal on Selected Areas in Communications (JSAC), a top academic journal in the field of computer network systems and ACM UbiComp 2022, a top international academic conference in the field of ubiquitous computing; one long article was published in ACM MobiCom 2022, the top international academic conference in the field of mobile computing, and won the second place in the Best Paper of the year Award Runner-up）
。

Ubiquitous wireless sensing refers to the technology of using ubiquitous wireless radio frequency signals in the environment to achieve non-contact perception of people and things in the
environment.
These wireless signals come from a variety of wireless communication devices deployed in our lives, including Wi-Fi, 4G/5G, RFID, and radar
.
Using these ubiquitous wireless signals, analyze the fluctuation pattern of wireless signals affected by human activities and extract signal characteristics, which can identify people's action behavior, monitor respiratory heartbeat, and sleep conditions; Or realize intrusion detection, indoor positioning, human trajectory tracking and other types of applications
.

As one of the most important signal characteristics of wireless sensing, Doppler speed can be extracted from the WiFi signal received by the receiving end, thereby supporting a series of applications
such as trajectory tracking and gesture recognition 。 However, when using WiFi signals to estimate the speed of a target's motion, there are two important theoretical questions that have not yet been answered: (1) Is the accuracy of the target's speed estimation consistently the same at any location within the WiFi sensing range? (2) Since the placement of WiFi devices is arbitrary in reality, how can accurate speed estimation of the target be achieved? The paper "Rethinking Doppler Effect for Accurate Velocity Estimation With Commodity WiFi Devices" explores and answers
the above questions 。 Starting from the basic concept of relativistic Doppler effect, based on the Fresnel zone perception model proposed by the team, Zhang Daqing's team established the relationship between Doppler frequency shift and target motion speed in WiFi perception system, and quantitatively analyzed how the speed (including speed size and direction) estimation accuracy in WiFi contactless perception is affected by factors such as the position of the target relative to the WiFi transceiver and the direction of target movement, and then gave the guiding principles
for achieving accurate speed estimation 。 Finally, using gesture recognition and human trajectory tracking as applications, the paper gives an example
of how to extract velocity features that are not affected by factors such as position and direction of movement.
The research results were published in the CCF Class A international journal IEEE JSAC 2022, and the first author of the paper was Dr.

Kai Niu of the team.

Figure 1.
Physical interpretation of the relationship between the Doppler shift and the velocity of target motion

In ubiquitous wireless sensing systems, the accuracy of sensing distance, range, and signal parameter estimation is often affected
by signal noise.
Zhang Daqing's team has made significant progress
in the basic theory and application of wireless sensing by defining the perceptual signal-to-noise ratio model and introducing it into the field of ubiquitous wireless perception.
Previous WiFi perception work lacked research and exploration of wireless sensing range, which made WiFi perception system lack theoretical guidance when deployed in real environments, and system performance was guaranteed
by trial and error in various scenarios 。 Through the introduction and research of the concept of "perceived signal-to-noise ratio", the paper "Placement Matters: Understanding the Effects of Device Placement for WiFi Sensing" gives the perceptual range boundary formula of ubiquitous wireless perception system for the first time in the world, revealing the law of changing the boundary shape of the sensing range with the distance of the transceiver device: that is, with the increase of the distance between WiFi devices, The range of perception changes from a small ovoid to a large ovate, then the ovate indents inward, and finally becomes two small ovates surrounding the transceiver (see Figure 2).

When the WiFi perception system is deployed in the real environment, only the location of the transceiver device needs to be guided according to the above theory, which can solve two common problems in WiFi perception: small sensing range and interference from distant targets
.
For example, by properly placing transmitters and receivers, the perceived range of a human tracking system can be increased by 200%.

By increasing the distance between the transmitter and receiver, the fine-grained breathing of the target can still be accurately perceived by the system even if the interferer is 0.
5 meters away from the sensing target
.
The relevant work was published in the CCF Class A international conference UbiComp 2022 (ACM IMWUT), and the first author of the paper is doctoral student Xuanzhi
Wang.

Figure 2.
Perceived coverage boundaries at different LoS lengths, where red and black circles are Tx and Rx device locations

In scenarios such as gesture recognition, the signal quality will change with different device placement, gesture position, movement direction, etc.
, so preprocessing the signal quality corresponding to the user's gesture action as constant and unified will cause the accuracy of gesture recognition to be unstable
.
In fact, when the hand is in different positions and orientations, sometimes the signal change is much greater than the noise, and sometimes the signal change is close to the noise, and the corresponding perceived signal-to-noise ratio and signal perceived quality are changed
.
Aiming at the problem of inconsistent signal perception quality of hand movement in different positions and downward direction, the paper "Towards Robust Gesture Recognition by Characterizing the Sensing Quality of WiFi Signals" proposes a signal perception quality measurement model
.
By characterizing the relative relationship between effective perception signals and noisy signals during hand movement, the perceived quality
of WiFi-CSI signals at different times and locations can be quantified in real time.
Based on the perceived quality of wireless signals, this paper further proposes a signal preprocessing framework
based on perception quality indicators.
Specifically, the framework classifies signals of different perceived qualities
.
For signals with good perception quality that can be used for identification tasks, the framework uses multi-carrier information to further improve signal quality.
For signals with poor perception quality that cannot be used to recognize tasks, the framework infers hand movement information
at this time through prior knowledge.
Finally, accurate Doppler velocity information can be extracted from hand movements in different positions and orientations, which can significantly improve the performance and robustness
of gesture recognition systems.
The work was published in the CCF Class A International Conference UbiComp 2022 (ACM IMWUT), and the first author of the paper is PhD student Ruiyang
Gao.

Figure 3.
Different hand activities have different signal quality, and a stable gesture recognition system can be constructed by measuring the signal quality

Based on the perceptual signal-to-noise ratio model, Zhang Daqing's team has also made significant progress
in improving the wireless sensing distance.
In the previous wireless sensing system, when the target was far away from the sensing device, the weak reflected signal was easily drowned out by the noise, making the perception system unable to work
.
In view of this common challenge, the paper "DiverSense: Maximizing Wi-Fi Sensing Range Leveraging Signal Diversity" proposes a sensing signal-to-noise ratio enhancement technology based on time-frequency spatial domain, which greatly improves the sensing distance of wireless signals and expands the sensing range
of WiFi 。 In WiFi perception systems, the channel state information (CSI) collected by the wireless device has a definite relationship with the motion of the target and the position relative to the wireless device, while the random noise roughly satisfies the Gaussian distribution
.
According to the law of large numbers, when a large number of noise-free signals are sampled independently and distributed, the signal converges to the desired value
.
Based on this principle, the WiFi signals collected in the time (multi-moment), frequency (multi-carrier), and space (multi-antenna) domains are used as multiple samples, and by aligning and fusing the signals, the signal-to-noise ratio can be greatly improved and the real human movement information
can be restored.
To this end, the team took the breath sensing application as an example to implement the DiverSense system based on commercial WiFi devices, advancing the human breathing detection distance from the current maximum of 11 meters to a new record
of 40 meters in the corridor environment.
The work of WiFi-based telebreathacity detection was published in the CCFA international conference UbiComp 2022, and the first author of the paper is doctoral student Yang
Li.

Figure 4.
CSI signal and extracted breathing waveform after 40 m breath perception experimental scene in corridor and enhanced signal-to-noise ratio

An important feature missing from existing wireless contactless sensing is in device movement scenarios, where perception cannot
be realized.
The reason why this task is challenging is that the signal changes caused by the irregular movement of the equipment are superimposed with the signal changes caused by the perceived target movement, and it is difficult to decompose
.
The paper "Mobi²Sense: Empowering Wireless Sensing with Mobility" proposes for the first time a new model and technology for non-contact wireless sensing in mobile scenarios, which stems from the miniaturization and low cost of radar chips in recent years, as well as the integration of various mobile devices such as handheld devices and mobile robots, which greatly enhances the urgent need
for wireless sensing in mobile scenarios.
The Mobi²Sense mobile sensing system proposed by Zhang Daqing's team cleverly uses the reflection signal of static objects in the environment to depict the movement law of the equipment, so as to select the reflected signal of the static reference object as the reference signal; Then the received signal and the reference signal are divided to eliminate the device motion components in the received signal to accurately restore the original motion signal
of the target.
A large number of experiments have shown that Mobi²Sense can capture subtle perceptual target movements in moving scenes, "hear" the sound vibrations of the speaker to restore sound, "see" the breathing state of the human body, and "recognize" multi-target gestures
with high accuracy.
The system can be used in emergency scenarios, doctors hold radar equipment to sense patients' vital signs, and use mobile robots to track and monitor the health status of the elderly at home, greatly broadening the application scenarios
of wireless sensing.
The research was presented at ACM Mobicom 2022, the top conference for mobile computing systems, and won the Best Paper Award Runner-up
.
This work was jointly completed by Peking University, the Institute of Software of the Chinese Academy of Sciences and the University of Massachusetts, and the first author Zhang Fusang is an associate researcher at the Institute of Software of the Chinese Academy of Sciences (previously a postdoctoral fellow at Peking University).

Figure 5.
The Mobi²Sense mobile sensing system can be used on a handheld device or mounted on a mobile robot platform to accurately sense target activity

Introduction of the corresponding author Zhang Daqing

Zhang Daqing is a chair professor at the School of Computer Science, Peking University, an academician of the European Academy of Sciences, an IEEE Fellow, and the director of
the CCF Pervasive Computing Committee 。 Since joining Peking University in 2014, he has been committed to the research of ubiquitous wireless perception theory and technology, and his team proposed the Fresnel region-based wireless perception theory for the first time in the world, revealing the relationship between the activity location, orientation, size, location of WiFi transceiver devices and wireless receiving signals of sensing targets.
A series of important properties of CSI quotient model, velocity-Doppler velocity model and wireless signal propagation in indoor environment are proposed.
By introducing the concept of perceived signal-to-noise ratio, the basic theoretical problems in the fields of WiFi perception limit, perception boundary, and perceived signal quality are answered, which provides a new theoretical basis for wireless perception based on WiFi, 4G/5G signal and radar signal.
It also achieves the best international performance
in a series of sensing applications such as contactless respiratory monitoring, indoor positioning, activity trajectory tracking and intrusion detection.

Since 2016, Zhang Daqing's team has published more than 30 papers at the ACM UbiComp Conference, a top international academic conference in the field of ubiquitous computing, and the cumulative number of papers published has always ranked first
in the world.
Among them, in the four years from 2016 to 2019, 4 long articles on wireless perception were cited in the top 3 of the UbiComp conference that year; In the past two years, he has won the ACM UbiComp 2021 Distinguished Paper Award and the second place of the ACM MobiCom 2022 Best Paper Award Runner-up
.
In 2022, Zhang Daqing's team, Huawei and other enterprises cooperated to take the lead in commercializing WiFi sensing in China, which shows that Peking University's research level in the field of pervasive computing and wireless sensing continues to be at the forefront of the world
.