Abstract — In real world applications, a multichannel acquisition system is susceptible of having one or many of its sensors displaced or detached, leading therefore to the loss or corruption of the recorded signals. In this paper, we present a technique for detecting missing or corrupted signals in multichannel recordings. Our approach is based on Higher Order Statistics (HOS) analysis. Our approach is tested on real uterine electromyogram (EMG) signals recorded by 4x4 electrode grid. Results have shown that HOS descriptors can discriminate between the two classes of signals (missing vs. non- missing). These results are supported by statistical analysis using the t-test which indicated good statistical significance of 95% confidence level. I. INTRODUCTION lectrophysiological signals are believed to carry detailed information relative to the system that generates them [1]. Over the years, these signals have been the subject of extended research in a variety of directions including signal acquisition equipment to areas such as digital signal processing, and pattern recognition. The analysis of these electrophysiological signals has completely changed the way various diseases previously were diagnosed in clinical routine. As a result, many of the problems confronting health professionals can be solved today by analyzing these signals recorded noninvasively from the patients [1]. The non-invasive acquisition of an electrophysiological signal is usually done through the attachment of a minimum of two electrodes to the body surface. However, multiple electrode configurations are commonly used in clinical practice to obtain a spatial description of the underlying bioelectrical phenomenon. The major applications in non- invasive multichannel electrophysiological signal acquisition would be those of the electroencephalogram (EEG) [2, 3], the electrocardiogram (ECG) signals [4] and the uterine electromyogram [5]. Studies have shown that, although there is a correlation between the electrical activities recorded at different sites, the characteristics of the recorded signal Manuscript received March 29, 2012. This work was achieved in collaboration between Rafik Hariri University (Lebanon), University of Technology of Compiègne (France) and the Lebanese University (Lebanon). Ramzi Halabi, Mohamad O. Diab and Bassam Moslem are with Rafik Hariri University (RHU), College of Engineering, Bio-instrumentation Department, Meshref, Lebanon.(Corresponding author: +961 5 601 381, email: halabiro@students.hcu.edu.lb). Catherine Marque is with the University of Technology of Compiègne (UTC), CNRS UMR 6600 Compiègne, Cedex, France. Mohamad Khalil is with the Lebanese University, AZM center for the research in biotechnology, Tripoli - Lebanon depend on the position of the recording electrode [1]. With the use of multiple electrodes, it has become possible to look deeper into the electrophysiological activity of a system, evaluate the correlation between signals recorded from spatially distributed regions and characterize it by new features such as the synchronization and the propagation velocity. Moreover, it was shown that the joint use of spatial data from different spatial sensors has the potential of solving difficult pattern recognition problems including the classification of electrophysiological signals [6]. However, in clinical practice, continuous recording of all channels is often cumbersome especially when a continuous monitoring for prolonged durations is to be performed. As a result, one or multiple signals can be lost, masked by noise or even corrupted during an ongoing recording session due to electrode displacement issues. In fact, there is always risk of electrode drift, misplacement, displacement or even complete detachment, leading to the corruption of the desired data and therefore misdiagnosis. Herein, corrupted signals of all sorts are termed missing signals for the sake of generalization. Hence, the detection of missing signals in multichannel recordings represents a crucial objective of biomedical signal processing so as to mitigate the influence of the misleading results of their analysis in biomedical research. In this paper, we present an approach for detecting missing signals in multichannel recording by using higher order statistics (HOS) descriptors since these descriptors are known to be sensitive to the shape variation of the amplitude distribution of a signal. Indeed, shape variation is expected to occur following the loss or the corruption of the recorded signal. Our approach is then tested on real uterine electromyogram (EMG) signals recorded by a matrix of 16 electrodes placed on the abdominal wall of pregnant women. Herein, we use the absolute values of kurtosis and skewness derived from the recorded channels in order to discriminate between contraction signals and missing signals found in the database. The values of the above stated parameters are compared between the 2 classes (non-missing vs. missing signals). Finally, the obtained results are discussed. II. MATERIALS AND METHODS A. Uterine Electromyogram (EMG): Uterine EMG signal, also called electrohysterogram (EHG), is the bioelectrical signal recorded noninvasively from the abdominal wall of pregnant women during their gestational period. It represents the noninvasive space-time recording of the uterine electrical activity with one or multiple independent sensors that capture some aspect of Detecting Missing Signals in Multichannel Recordings by Using Higher Order Statistics Halabi R., Diab M.O., Moslem B., Khalil M. and Marque C. E 34th Annual International Conference of the IEEE EMBS San Diego, California USA, 28 August - 1 September, 2012 3110 978-1-4577-1787-1/12/$26.00 ©2012 IEEE