IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 50, NO. 4, JULY 2001 981 Mobile Speed Estimation for TDMA-Based Hierarchical Cellular Systems Chengshan Xiao, Karl D. Mann, and Jan C. Olivier Abstract—In this paper, a new algorithm is presented for esti- mating mobile speed for handoff in hierarchical cellular systems. The proposed algorithm is based on normalized autocorrelation values of received signals to estimate mobile speed; it contains six steps. First, the instantaneous power of the received baseband signal is calculated to remove the frequency offset and data/speech information-bearing signals, while keeping the Doppler frequency information. Second, the calculated power signal is filtered using a low-pass linear phase finite impulse response filter to suppress interference and noise. Third, the filtered power signal is deci- mated to ease the computational burden, while the decimation factor is properly chosen to avoid frequency aliasing. Fourth, autocorrelation values of the decimated filtered power signals are calculated on shifting slot by slot to suppress the “slot burst frequency” interference. Fifth, the calculated autocorrelation values are normalized to suppress the power fluctuation of the received signals. Finally, the normalized autocorrelation values are compared with thresholds to estimate mobile speed. The simulation results indicate that the new algorithm works very well for both nondispersive channels and dispersive channels to distinguish fast and slow moving mobiles. The method has very low latency, with results being available typically within 1 s after communication is established, and it can report estimation result every second or less. The algorithm has been implemented by software code into Nortel’s base-station radios and tested in Nortel’s wireless communications labs. The lab test results are very close to the computer simulation results, which have very good estimation accuracy. Index Terms—Digital cellular systems, hierarchical cellular architecture, macrocell and microcell overlaying system, mobile speed estimation. I. INTRODUCTION T O COPE with the rapidly increasing demand for mobile and personal communications, hierarchical cellular sys- tems, which have multiple-layer cellular cells, are being de- ployed in dense urban areas [1]. Specifically, a two-layer hi- erarchical cellular system consists of microcells overlaid with macrocells, where a macrocell is the union of many microcells. Thus slow-moving mobiles are assigned to microcells and fast- moving mobiles are assigned to macrocells. This approach has an objective of decreasing the handoff rate for fast-moving mo- biles. Hence, a reliable mobile speed estimator is desirable. The Manuscript received February 8, 1999; revised June 6, 2000. This paper was presented in part at the IEEE Vehicular Technology Conference (VTC’99), Am- sterdam, The Netherlands, September 1999. C. Xiao is with the Department of Electrical Engineering, University of Mis- souri, Columbia, MO 65211 USA (e-mail: xiaoc@missouri.edu). K. D. Mann is with Wireless Solutions, Nortel Networks, Ottawa, ON K2G 6J8, Canada (e-mail: kmann@nortelnetworks.com). J. C. Olivier is with the Nokia Research Center, Dallas, TX 75039 USA (e-mail: Jan.Olivier@nokia.com). Publisher Item Identifier S 0018-9545(01)07006-2. benefits of decreasing handoff rate include an increase in ca- pacity for the system and a decrease in the number of dropped calls. As well, voice quality is improved due to a reduction of the number of times this voice is muted for handoff. In the literature, there are a few methods of mobile speed es- timation that have been published [2]–[10]. Specifically, in [2], the estimation of maximum Doppler frequency can be used to estimate mobile speed. This approach will not be reliable if the spectrum of a fading process has multiple local maxima. The maximum Doppler frequency cannot be easily determined due to different levels of energy at these frequencies. In [3], the di- versity switching number is used to estimate mobile speed, but it is pointed out in [4] that this method is highly dependent on the fading distribution statistical properties (Rayleigh fading, Ri- cian fading, etc.), and since it is not easy to know the distribution in advance, this method is not very reliable in practice. In [5], based on deviation of received signal strength, two methods are proposed to estimate mobile speed for GSM radios. The first one works fine when the channel has no intersymbol interfer- ence, and it fails in the presence of intersymbol interference, i.e., dispersive channels. The second method of [5] uses pattern recognition to overcome the limitations of the first method on dispersive channels but leads to high computational complexity and may not be reliable due to the nature of pattern recognition of dispersive channels. The two methods in [5] can also be ap- plied for IS-136 time-division multiple-access(TDMA) radios. In [6], multiple base-station and multidimensional scaling are used to estimate mobile speed; this method may be expensive in practice. In [7], the level crossing rate (LCR) is used to estimate mobile speed for Advanced Mobile Phone System (AMPS) ra- dios, but unfortunately it is not reliable for IS-136 TDMA radios because the 50-Hz interference, which is caused by the TDMA slot burst repeating time 20 ms, is inside the Doppler frequency range: 1) 0–80 Hz for 850-MHz radios with maximum mobile speed of 100 km/h; 2) 0–180 Hz for 1900-MHz radios with maximum mobile speed of 100 km/h. During the review of this paper, another new method was pro- posed in [10] by using wavelets to estimate mobile speed. In this paper, a new and mathematically simple algorithm is pre- sented for estimating mobile speed. The proposed algorithm is based on normalized autocorrelation values of received signals. In principle, the idea of this algorithm is applicable to all the cellular systems with angle-modulated signals including FM, 4DQPSK, GMSK, 8PSK, and QPSK modulations for AMPS, IS-136 TDMA, GSM, EDGE, and code-division multiple-ac- cess(CDMA) systems, respectively. However, in this paper, we 0018–9545/01$10.00 ©2001 IEEE