Low Complexity Frequency Domain TOA Estimation for IR-UWB Communications Monica Navarro ∗ , Simon Prior ∗‡ and Montse Najar ∗† ∗ Centre Tecnologic de Telecomunicacions de Catalunya Castelldefels, Barcelona, Spain {monica.navarro,simon.prior,montse.najar}@cttc.es † TSC - Universitat Politecnica de Catalunya ‡ TU Graz Graz, Austria Abstract— This paper addresses the problem of TOA estima- tion for ranging applications in IR-UWB. It considers a low complexity frequency-domain approach for the estimation of the TOA that allows sub-Nyquist sampling rate receivers. Two diversity schemes are investigated in combination with the TOA estimation algorithm and are evaluated over realistic channel models, yielding accurate ranging estimation. I. I NTRODUCTION Communication systems based on impulse radio ultra- wideband (IR-UWB) have been envisaged as radio communi- cation systems that could enable very accurate ranging and lo- cation applications, given the extremely short duration pulses. This high time resolution nature of the UWB signal makes TOA estimation method a good candidate for positioning estimation in UWB communications. Ranging accuracy depends on how precisely the receiver can discriminate the first arriving signal, which in a multipath environment may not be the strongest. In the literature most of ranging techniques are based on time-domain TOA estimation methods. The maximum likelihood (ML) solution has practical limitations due to the requirement of very high sampling rates. Typically, the conventional correlation-based approach results in a very slow TOA estimator [1] requiring an exhaustive search over a large number of beams. Iterative ML approaches have been also studied [2] but yet requiring very high rate sampling. Recently, it has appeared in the literature proposals to reduce the sampling constraints and time intervals required for esti- mation of time-domain based TOA estimators. An approach addressed by several authors consist on a two step TOA estimation process consisting of an initial coarse estimation of the TOA followed by a higher resolution stage. Following this strategy, the authors in [3] propose a first rough TOA estimate based on the received signal energy, followed by a low-rate correlation stage that estimates the TOA based on hypothesis testing. In [4] a similar two step estimator is proposed based on a threshold based energy detection receiver. The scheme allows for a symbol rate sampling but requires using several symbols and an appropriate design of the signal waveform. Two stage approach is also considered in [5] were a low cost non-coherent receiver based on an energy detection stage is proposed. The basic principle is based on integration windows which time resolution changes between the two stages. A critical parameter for these type of estimators lies on the threshold selection. Motivated by recent work on UWB receiver architectures that provides direct samples of the received signal in the frequency domain at sub-Nyquist sampling rate [6] [7], in this paper we propose the use of low complexity frequency-based TOA estimation techniques. The use of frequency domain TOA estimation methods has recently been considered to provide high resolution estimates for positioning applications in UMTS framework [8]. In [9] the authors also examined frequency domain high-resolution methods for TOA estimation for indoor geolocation. The aim of this paper is to provide a low complexity TOA estimation method requiring a sub- Nyquist sampling rate. The rest of the paper is organized as follows. The signal model is presented in Section II. Section III introduces the frequency-domain TOA estimation algorithm followed by the proposed ranging technique in Section IV. Evaluation and per- formance results are detailed in Section V. Finally, conclusions are drawn in Section VI. II. UWB SIGNAL MODEL We considered an IR-UWB system where transmission of an information symbol is typically implemented by the repetition of N p pulses of very short duration, s(t)= Np-1  p=0  E p p(t - pT pr ) where E p denotes the energy of one pulse, p(t) refers to the single pulse waveform, typically a Gaussian monocycle or its derivatives 1 , and T pr is the repetition pulse period. 1 Modulation and time hopping can be explicitly include in the pulse waveform such us p(t)= ps(t - b i T δ - c p i Tc) with ps(t) being typically a Gaussian monocycle or its derivatives, b i the information symbol, T δ modulation time shift, c p i the time hopping sequence and Tc the chip interval. We assume symbol synchronization is available at the receiver (eg: via transmission of pilot symbols) 1-4244-0063-5/06/$2000 (c) 2006 IEEE