0018-9294 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TBME.2017.2718179, IEEE Transactions on Biomedical Engineering TBME-00633-2017 Abstract—In this paper we propose a fast novel non-linear filtering method named Relative-Energy (Rel-En), for robust short-term event extraction from biomedical signals. We developed an algorithm that extracts short- and long-term energies in a signal and provides a coefficient vector with which the signal is multiplied, heightening events of interest. This algorithm is thoroughly assessed on benchmark datasets in three different biomedical applications namely, ECG QRS-complex detection, EEG K-complex detection, and imaging photoplethysmography (iPPG) peak detection. Rel-En successfully identified the events in these settings. Compared to the state-of-the-art, better or comparable results were obtained on QRS-complex and K-complex detection. For iPPG peak detection, the proposed method was used as a preprocessing step to a fixed threshold algorithm that lead to a significant improvement in overall results. While easily defined and computed, Rel-En robustly extracted short-term events of interest. The proposed algorithm can be implemented by two filters and its parameters can be selected easily and intuitively. Furthermore, Rel-En algorithm can be used in other biomedical signal processing applications where a need of short-term event extraction is present. Index Terms—Non-linear signal processing, impulse detection, short-term event extraction, QRS-complex detection, biomedical signal processing. I. INTRODUCTION IOMEDICAL signal processing is the study of the measurements recorded by physiological instruments. Analysis of these measurements provides physicians with important physiological information that may help them to uncover underlying dynamics of human health. It also allows them to determine patient health state and to choose the right treatment. Physiological signals comprise different waveforms, the extraction of which often being the first step in their investigation. Over the last decades, several spike/waveform extraction methods have been proposed in the literature. Generally, the signal is preprocessed by means of high-, low-, or band-pass filtering, followed by a comparison against detection logics and thresholds, to determine the veracity of the peaks [1]. A simple approach is to apply a threshold on the raw or the absolute value of a signal to extract impulsive * S. Yazdani is with the Applied Signal Processing Group, Department of Electrical Engineering, Swiss Federal Institute of Technology, Lausanne, Switzerland (correspondence e-mail: sasan.yazdani@epfl.ch). S. Fallet is with the Applied Signal Processing Group, Department of Electrical Engineering, Swiss Federal Institute of Technology, Lausanne, Switzerland (e-mail: sibylle.fallet@epfl.ch). J.-M. Vesin is head of the Applied Signal Processing Group, Department of Electrical Engineering, Swiss Federal Institute of Technology, Lausanne, Switzerland (e-mail: jean-marc.vesin@epfl.ch). events, as proposed for neural recordings [2-4]. A common way to set this simple threshold is to base it on an estimate of the standard deviation of the noise [2-3]. The Plosion index (PI), is another method used to detect impulse-like events in signals [5]. This index is defined as, () = |()| (∑ |()| + 2 =− 1 )/( 2 + 1 +1) (1) Where () denotes the  ℎ sample of the input signal . Parameters  1 and  2 describe the interval on which the average is applied. PI has been used in speech processing to detect closure burst transitions of stops and affricates [5]. This index has also been used to extract R-waves from ECGs [6]. The slope characteristics of the input signal have proven to be a good feature for short-term event extraction. The idea is to scrutinize the first derivative of the input signal in order to detect abrupt changes in the amplitude, which represent the onset of action potentials or impulsive waveforms [7-11]. The nonlinear energy operator [12], also known as Teager energy, has also been used to detect impulses. This operator captures high frequency local activities using an input sample and its immediate neighbors. After applying this operator, the output is compared against a threshold and impulses are extracted [2,13]. The abovementioned methods are easy to implement and perform well in normal settings in which the input signal is not contaminated by perturbations. More elaborated yet robust methods such as template matching and wavelet transforms have been proposed to deal with noise-contaminated signals. Template matching methods scan the signal in order to find instances that are similar to a pre-defined set of templates. Usually the matching operation is carried out by studying a similarity measure such as the Euclidean distance between an instance and the pre-defined templates [7,14-16], or by calculating their cross-correlation [17-18]. Template matching algorithms generally perform better compared to the aforementioned ‘thresholding’ methods. Moreover, they can identify different morphologies and extract several types of waveform from the signal. However, these methods require a priori knowledge of the morphology of the waveform(s) of interest. Since these morphologies can be application- or even subject-dependent, they need to be studied and might even require expert input in order to be extracted. The wavelet transform is another interesting approach for waveform extraction [19]. By offering good frequency resolution at low frequencies and high time resolution at high frequencies, wavelet-based methods are vastly popular in automatic waveform detection [3,19-20]. Furthermore, these A Novel Short-term Event Extraction Algorithm for Biomedical Signals Sasan Yazdani*, Sibylle Fallet, and Jean-Marc Vesin, Member, IEEE B