On the use of instantaneous mean frequency estimated from the Hilbert spectrum of facial electromyography for differentiating core affects

Abstract

Facial expression comprised of single or combination of facial action units (AUs) is one of the most important communication channels transferring affective states in our social life. Electromyography (EMG) activity captured over specific facial muscles could be used for interpreting one's affective experience. The present study investigated the effects of facial EMG changes in mean instantaneous mean frequency (IMNF) captured over corrugator supercilii and zygomaticus major for differentiating core affects. Firstly, an adaptive filter has been employed to deal with power line interference in the preprocessing of EMG signal. Secondly, a novel method -IMNF estimated via Hilbert spectrum was applied, which enables us to represent the characteristic of facial EMG in joint timefrequency domain. Thirdly, mean IMNF changes of facial EMG activity elicited by affective picture blocks were computed for differentiating core affects which describe affect in valence and arousal. The results indicated that low valence/high arousal affective state can be discriminated by corrugator supercilii EMG from neutral valence/low arousal, high valence/high arousal, high valence/low arousal and low valence/low arousal affective conditions, in which corrugator supercilii EMG in response to low valence/high arousal stimuli was greater than others. In contrast, zygomaticus major EMG indicated high valence/high arousal affective state with the highest response, compared to other four affective states; Zygomaticus major EMG also revealed significant differences between low valence/low arousal and high valence/low arousal, low valence/low arousal and low valence/high arousal.

Classificação de atividade eletromiográfia facial de indivíduos saudáveis e com hanseníase por meio de máquina de vetores de suporte

“…As características mais utilizadas na literatura são RMS (valor médio quadrático) e MAV (valor médio absoluto), podendo desta última se extrair outras características com a primeira e a segunda derivada. No domínio da frequência podemos analisar o espectro de potência do sinal, o que permite quantificar características como frequência mediana,frequência média e frequência modal(Tan et al, 2012).…”