Affiliation:
1. Department of Instrumentation and Control Engineering, Dr. BR Ambedkar National Institute of Technology,
Jalandhar, Punjab, India
Abstract
Background::
Facial emotion recognition (FER) technology is enumerated as a productive
interface in several operations, which has been specifically focused on as a substitute
communication path among a user and an appliance for human computer interface in the previous
decade. The efficiency of the facial identification model straightaway relies on the capability
of classification methods. In addition, an appropriate swap between recognition efficiency
and computational cost is reckoned as the most important factor for planning such models.
Methods::
The objective of this paper was to classify the facial emotion electromyogram (EMG)
signals by means of a neural network algorithm (NN), support vector machine (SVM) algorithm,
and Naive-Bayes algorithm. This research work was directed towards the correlation among the
classification accuracies by applying distinct feature extraction procedures on fEMGs. At first,
eight participants (six male and two female) were recruited for data recording. Four electrodes
were placed on each participant's face for capturing facial gestures (happy, angry, sad, and fear)
and two electrodes were placed on the wrist for grounding purposes. Data were recorded by using
BIOPAC MP150. After this, the signals were filtered using a band-pass filter and segmentation
techniques for enhanced processing. After that, the time-domain and frequency-domain feature
extraction procedures were carried out. Time domain and frequency domain features were
applied to recorded signals. In this research, we used LabVIEW and MATLAB to produce a set
of characteristics from fEMG signals for four emotional conditions, such as anger, sad, fear, and
happy. After the feature extraction process, the extracted features were aligned into respective
emotions by applying classifiers. The extracted features were further trained and classified by
applying the SVM classifier, neural network classifier, and Naive Bayes classifier in MATLAB
2020.
Results::
The SVM classifier and neural network classifier generated an accuracy of 93.80% and
96.90%, respectively, whereas the Naive Bayes classifier generated an accuracy of 90.60%.
Conclusion::
Facial emotion recognition (FER) is foresighted as a progressive or futuristic model,
which has attracted the attention of researchers in several areas of learning due to its higher
prospects in distinct applications. Acknowledgment of the emotions through biomedical signals
produced from movements of facial muscles is lately presented using an explicit and authentic
route.
Publisher
Bentham Science Publishers Ltd.
Reference64 articles.
1. Zecca M.; Micera S.; Carrozza M.C.; Dario P.; Control of multifunctional prosthetic hands by processing the electromyographic signal. Crit Rev Biomed Eng 2002,30(4-6),459-485
2. Veer K.; ‘Interpretation of surface electromyograms to characterize arm movement’. Instrum Sci Technol 2014,42(5),513-521
3. Doulah A.B.M.S.U.; Fattah S.A.; Wavelet domain feature extraction scheme based on dominant motor unit action potential of EMG signal for neuromuscular disease classification Proceedings of the Eighteenth International Conference on Artificial Intelligence and Statistics AISTATS 2015, IEEE Trans Biomed Eng, deeply supervised nets. 8
4. Veer K.; A technique for classification and decomposition of muscle signal for control of myoelectric prostheses based on wavelet statistical classifier. Measurement 2015,60,283-291
5. Dwyer R.; Detection of non-Gaussian signals by frequency domain kurtosis estimation International Conference on Acoustics, Speech, and Signal Processing (ICASSP’83) ,8,607-10