Facial Gesture Recognition Using Two-Channel Bio-Sensors Configuration and Fuzzy Classifier: A Pilot Study Mahyar Hamedi 1- Department of Biomedical Eng., Science and Research Branch, Islamic Azad University Tehran, Iran 2- Faculty of Biomedical and Health Science Engineering, Universiti Teknologi Malaysia Skudai, Malaysia Hamedi.mahyar@gmail.com Iman Mohammad Rezazadeh Department of Biomedical Eng., Science and Research Branch, Islamic Azad University Tehran, Iran i.rezazadeh@srbiau.ac.ir Mohammad Firoozabadi 1- Faculty of Medical Sciences Tarbiat Modares University 2- Department of Biomedical Eng., Science and Research Branch, Islamic Azad University Tehran, Iran pourmir@modares.ac.ir Abstract—Facial gesture recognition has become an important issue in diagnostic, medical and industrial fields. Automatic recognition of facial gestures could be considered as an important factor in human-machine interface applications. Facial gesture recognition based on surface electromyography (SEMG) has been well thought-out in the recent decade. SEMG has accurate rates for facial gesture recognition since it records the electrical potential from facial muscles. This paper presents a method for recognizing 5 different facial gestures based on forehead two-channels bioelectric-signals. The recorded signals were processed in four steps: filtration, feature extraction (RMS), thresholding, and classification. The extracted features were classified into 5 facial gesture classes (rest, smile, frown, rage, and gesturing ‘notch’ by pulling up the eyebrows) by utilizing Fuzzy C-Means (FCM) classifier. Finally 90.8% recognition ratio has been achieved by applying our method on 4 subjects. Keywords-- Facial gesture recognition, SEMG, Fuzzy C- Means (FCM) I. INTRODUCTION In recent decades, there are many efforts done in the field of facial gesture and expression recognition. That is due to the fact that, face is one of the most visible and expressive of all the channels of communication such as emotions and facial gestures. On the other hand, it can play an important role wherever humans interact with machines. It is also one of the best sources of information in human’s body since many biopotential signals may be extracted from it (like EMG, EOG, EEG). Forehead bioelectric signals have potential to mirror other biosignals such as EEG or EOG accompanying with SEMG [1]. In addition, Facial gestures could convey non- verbal expressions, which plays an important role in interpersonal relations. There are many researches with different techniques which have been done in this area. Image processing is a popular and easy one with very low costs, feasible recognition (e.g. [2-4]). Surface electromyographic (EMGs) signals-based facial gesture recognition has been considered lately (e.g. [5, 6], [24]). Using bio-sensors mounted on facial muscles. This method has some privileges over other gesture recognition methods. It is strong against many environmental circumstances which are difficult to overcome by other methods of gesture recognition [8]. One of the most important issues is the number of bioelectric data channels (the number of electrodes). In previous projects which were done by other researchers, three channels to record their facial SEMG have been chosen [5, 6], [24]. That was because of the specific field of study and applications. There is a tradeoff between number of channels, user-friendliness of the interface, and accuracy of the method. There have been many attempts done on SEMG processing field to get better result in recognition ratio. In the part of feature extraction mostly used time domain features because of its specifications. These features are categorized into many methods such as MAV, SSI, VAR, ZC, and RMS which was used in many projects [5], [6], [9], [10], [24]. Furthermore, classification methods can be seen as another important fact in facial recognition using SEMG and many aspects have been employed for this approach (Neural Network, Fuzzy and Fuzzy Neural Network for example) (e.g. [5], [6], [11-13], [24]). International Conference on Electrical, Control and Computer Engineering Pahang, Malaysia, June 21-22, 2011 978-1-61284-228-8/11/$26.00 ©2011 IEEE 338