Blind Modulation Identification for MIMO Systems K. Hassan, C. Nsiala Nzéza, M. Berbineau Univ Lille Nord de France, F-59000 Lille, INRETS, LEOST, F-59650 Villeneuve d’Ascq kais.hassan@inrets.fr W. Hamouda Concordia University, Montreal, Quebec, H3G 1M8, Canada hamouda@ece.concordia.ca I. Dayoub Univ Lille Nord de France, F-59000 Lille, IEMN, DOAE, F-59313 Valenciennes Iyad.dayoub@univ-valenciennes.fr Abstract—Modulation type is one of the most important characteristics used in signal waveform identification and classi- fication. In this paper, an algorithm for blind digital modulation identification for multiple-input multiple-output (MIMO) systems is proposed. The suggested algorithm is verified using higher order statistical moments and cumulants of the received signal. A multi-layer neural network trained with resilient backpropa- gation learning algorithm is proposed as a classifier. The purpose is to discriminate among different M-ary shift keying linear modulation types and the modulation order without any priori signal information. This study covers different MIMO systems with and without channel state information (CSI). The proposed classifier is evaluated through the probability of identification where we show that our proposed algorithm is capable of identifying the modulation scheme with high accuracy in excellent signal-to-noise ratio (SNR) range. Index Terms—Higher order statistics, multiple-input multiple- output, modulation identification, neural networks, spatial mul- tiplexing, space-time block coding. I. I NTRODUCTION Over the last decade Multiple-Input Multiple-Output (MIMO) systems have shown a great importance as they provide a reliable and high-data-rate wireless communica- tions. Nowadays, MIMO is considered one of the promising technologies for developing the next generation of wireless systems. Recently, blind algorithms and techniques for MIMO signals interception have gained more attention. One essential step in the signal interception process is to blindly identify the modulation scheme of MIMO signals. Modulation identification has its roots in military ap- plications such as; communication intelligence (COMINT), electronic support measures (ESM), spectrum surveillance, threat evaluation and interference identification. Also recent and rapid developments in software defined radio (SDR) in the context of cognitive radio (CR) have given modulation identification more prominence in civil applications. Many modulation identification algorithms have been devel- oped for Single-Input Single-Output (SISO) systems [1]. These algorithms are generally divided into two categories. The first category is based on decision theoretic approach while the second on pattern recognition. The decision theoretic approach is a probabilistic solution based on a priori knowledge of probability functions and certain hypotheses [2], [3]. On the other hand, the pattern recognition approach is based on extracting some basic characteristics of the received signal called features [4]–[10]. This approach is generally divided into two subsystems: the features extraction subsystem and the classifier subsystem. However, the second approach is more robust and easier to implement if the proper features set is chosen. In the past, much work has been conducted on modulation identification [4]–[10]. The identification techniques, which have been employed to extract the signal features necessary for digital modulation identification, include spectral based features set [4], [5], higher order statistics (HOS) [6], [7], constellation shape [8], and wavelets transforms [9], [10]. With their efficient performance in pattern recognition problems (e.g., modulation classification), many studies have proposed the application of artificial neural networks (ANNs) as classi- fiers [5], [10]. In [11], Swami et. al proposed a simple yet very low complexity method, based on elementary fourth-order cumu- lants for the classification of digital modulation schemes. The robustness of this approach comes about not only from the resistance of HOS to additive colored Gaussian noise, but also from a natural robustness to constellation rotation and phase jitter. Also additive non-Gaussian noise can be handled if its fourth-order cumulants are known, or via simple preprocessing. So far, few researches have considered the application of modulation identification techniques in MIMO systems. For instance, Choqueuse et. al [12] adopted a maximum likelihood approach for the blind recognition of the modulation for MIMO systems using spatial multiplexing (SM). In their work the authors proposed two Likelihood based classifiers. The first one, called Average Likelihood Ratio Tests (ALRT), is optimal in the Bayesian sense but requires the knowledge of the chan- nel matrix. The second classifier, called Hybrid Likelihood Ratio Tests (HLRT), approximates the ALRT by replacing the channel matrix with its estimate. The major drawbacks of these methods are the high computational complexity and its need of perfect knowledge of the noise variance at the receiver side. In this paper we introduce a pattern recognition approach for modulation identification in MIMO systems. In our approach, the features extraction subsystem is based on the higher order cumulants (HOC) and the higher order moments (HOM) of the received signal. Our proposed classifier is a multi-layer artificial neural network trained using the resilient backpropa- gation learning algorithm (RPROP). Modulation identification is performed without any priori information of the received signal (e.g. probability functions, noise statistics, etc.). Then, our proposed algorithm is considered as semi-blind when 978-1-4244-5637-6/10/$26.00 ©2010 IEEE This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2010 proceedings.