Estimation of the Velocity of a Walking Person in Non-Stationary Indoor Environments from the Received RF Signal Rym Hicheri Faculty of Engineering and Science University of Agder NO-4898 Grimstad, Norway rym.hicheri@uia.no Matthias P¨ atzold Faculty of Engineering and Science University of Agder NO-4898 Grimstad, Norway matthias.paetzold@uia.no N´ eji Youssef Ecole Sup´ erieure des Communications de Tunis Universit´ e de Carthage 2083 El Ghazala, Tunisia neji.youssef@supcom.rnu.tn Abstract—Accurate estimation of the time-variant (TV) ve- locity, i.e., TV speed and TV direction of motion, of walking persons in indoor environment is of great importance in a variety of wireless indoor applications. This paper presents a novel method for estimating the velocity of a walking person in three-dimensional indoor environments, which are assumed to be equipped with a distributed 3 × 3 multiple-input multiple- output (MIMO) system. The approach estimates the TV speed, TV vertical angle-of-motion (VAOM), and TV horizontal angle- of-motion (HAOM) by fitting the spectrogram of the complex channel gain of a non-stationary indoor channel model to the spectrogram obtained from the received radio signals. The valid- ity of the estimation algorithm has been confirmed by comparing the estimated parameters of interest with their corresponding exact values, known from generated test signals. I. I NTRODUCTION The precise estimation of the time-variant (TV) velocity of moving objects/persons plays a key role in enabling a variety of mobile radio communications and indoor wireless applications. In the context of mobile radio communications, precise estimation of the mobile stations’ speed is required for the design of hand-off, adaptive modulation, and power con- trol algorithms [1]. The developed estimators can be classi- fied according to the statistical property they employ: level- crossings, covariance, maximum likelihood, and power spec- trum. Crossing-based methods calculate the number of level- crossings of the envelope and/or phase of the received signal, which is proportional to the speed of the mobile station [2], [3]. Covariance-based approaches determine the speed of the mobile units by means of the maximum Doppler frequency information which can be extracted from the covariance of the received signal [4], [5]. It should be noted that crossing- and covariance-based estimation techniques require the a priori knowledge of the statistical properties of the fading channel. Due to the limited studies on non-stationary channel statistics, the applicability of these methods is limited to wide-sense stationary (WSS) channels. Maximum likelihood-based speed estimation techniques require the knowledge of the signal- to-noise ratio and are complex to implement [6], [7]. Power spectrum-based methods have been developed based on the differentiation and derivation of the power spectrum of the fading channel [8]–[10]. Although these approaches have been extended to non-isotropic scattering [10], their applicability is limited to WSS channels, for which the speed is assumed to be constant. Additionally, the aforementioned speed estimation procedures do not provide any information regarding the direction of motion of the mobile units. In recent years, much attention has been paid to velocity- based human in-home activity tracking because of its key role in various wireless indoor applications, such as remote medical care, healthcare services, security and surveillance, and human-machine interaction [11]–[13]. Context-aware and wearable velocity-based devices can be classified according to the technique they are employing: video (image pro- cessing), sensor, and radio frequency (RF). Video-based in- home monitoring estimates the motion profile of moving/fixed objects/persons over time using a camera by detecting their boundaries [14]–[16]. Sensor-based human activity tracking in indoor environments can take the form of on-body or off-body devices which are equipped with sensors such as accelerometers, gyroscopes, and global positioning systems [17]–[20]. The RF-based motion tracking approach exploits the effects of a moving object/person on a received RF signal. The usefulness of this technique has been demonstrated by the Emerald [21], the mmVital [22], and the WiGait [23] devices. Interest in radar-based in-home activity tracking has also been growing due to its proven technology and safety [24]. As pointed out in [22], [25], and [24], RF-based approaches face several challenges, including their sensitivity to the rich scat- tering structure of indoor environments, i.e., fixed scatterers resulting from walls and furniture, and the TV direction of motion of the moving users. Our objective is to contribute to the topic of indoor human activity monitoring by proposing a new iterative procedure to estimate the TV velocity of a single moving person in a three-dimensional indoor environment that considers the effects of both fixed and moving objects. For simplicity, the walking person is represented by a single moving scatterer that may represent any part of the person’s body, e.g., the center of gravity, head or shoulder. The fixed scatterers represent the fixed objects in the room, such as walls, furniture, and decoration items. Here, the room is assumed to be equipped with a distributed 3×3 multiple-input multiple-output (MIMO) system. In the present work, the TV speed, TV vertical © 2018 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution. DOI: 10.1109/LATINCOM.2018.8613224