Double-Threshold Based Narrowband Signal Extraction J. Vartiainen, J. J. Lehtom¨ aki and H. Saarnisaari University of Oulu Centre for Wireless Communications (CWC) P.O. Box 4500, FI-90014 University of Oulu FINLAND Tel. +358-8-553 2967, Fax: +358-8-553 2845 E-mail: johanna.vartiainen@ee.oulu.fi Abstract—A localization algorithm based on double- thresholding (LAD) is a computationally simple method for localizing narrowband signals in the frequency domain. The method does not need any a priori information about the narrowband signal. The localization is based on two thresholds. The lower threshold is used to compose adjacent signal samples into clusters whereas the upper threshold is used to detect signals. The LAD can be applied in narrowband signal detection as well as in interference suppression. The simulation results show that the LAD gives quite good localization accuracy and the LAD is able to determine correct number of narrowband signals even over 95% of the cases. I. INTRODUCTION Overlay systems may become common in wireless communications due to the limited amount of frequency bands. In overlay systems, spread spectrum (SS) signals coexist with narrowband (NB) signals used in conventional communication systems. NB signals may also be interference, either natural or artificial. Detection and localization of those NB signals may lead to their mitigation [1] and, as a consequence, improvements in SS system’s performance. New technologies, such as cognitive radio [2], [3], use information about existing signals in order to achieve better use of radio frequencies. Dynamic frequency selection (DFS) [4] means that a free frequency band is searched for and selected before transmitting an information signal. Simple and efficient algorithms are needed for that purpose. A large number of interference excision or NB signal detection algorithms are based on the use of a threshold. The threshold can be calculated in numerous ways, see, e.g., [5], [6]. Threshold setting is a critical task since the performance of excision or detection algorithm depends on that. Even though the threshold is set properly there exist problems. One problem is that the NB signal energy at a certain frequency within the frequency band of the NB signal may temporarily This research was supported by the National Technology Agency of Finland, Nokia, the Finnish Defence Forces, Elektrobit and Instrumentointi. The work of J. Vartiainen was supported by the Finnish Cultural Foundation and Tauno T¨ onningin s¨ a¨ ati¨ o. drop below the threshold. This causes needless separation of the NB signal into two (or more) parts. Also, the noise (and a possible wideband signal) may temporarily yield to threshold crossing and cause falsely detected NB signals. In other words, if the threshold is high, the NB signal may be separated into several parts but false detections are avoided. On the other hand, if the threshold is small, false detections become more common but needless separation of the NB signal is avoided. Several methods have been proposed for reducing problems mentioned above, e.g., [7], [8], [9], [10]. However, many of these methods either have to have some a priori knowledge, e.g., from the noise level, or have high computational complexity. The localization algorithm based on double- thresholding (LAD) has been proposed in [11] by the authors. The LAD thresholds can be calculated with the simple and efficient forward consecutive mean excision algorithm (FCME) [6]. The LAD does not need a priori knowledge about the signal to be detected or the noise level, and its computational complexity is relatively low. Herein, the LAD is briefly presented and its performance is studied by means of the correct number of narrowband signals it is able to detect as well as the bit error rate (BER). The latter has not been studied previously. II. SYSTEM MODEL The received signal is assumed to have the form r(n)= s(n)+ i(n)+ w(n), (1) where s(n) is caused by the direct sequence (DS) SS signal, i(n) is a narrowband signal and w(n) is the noise process. Two types of narrowband signals are considered; these are sinusoids and band-limited binary phase shift keying (BPSK) signals. 0-7803-8887-9/05/$20.00 (c)2005 IEEE