Research On Effect Mechanism Of Service Robots’ Anthropomorphic Visual Features On User Acceptance

With the decrease of growth rate of the global population,the problems of aging and labor shortage are becoming more and more serious.Service robots are gradually becoming labor substitutes.Artificial intelligence and Internet of Things technology provide strong development potential for service robots,and the upgrading of consumer demand under the background of Industry 4.0 has created a broad market prospect.For a long time,improving the functional design and the material structure are the focus of improving the user acceptance of service robots.But current researches ignore the significant influence of emotion,there are few studies analyzing the effect mechanism of service robots’ anthropomorphic visual features on user acceptance from the perspective of emotional experience.Through multimodal measurement experiment and ERP measurement experiment,this paper selects emotion measurement indexes to study the impact of different levels of anthropomorphic visual features on user acceptance.The specific research contents and results are listed as follows:(1)Proposing a theoretical framework of user acceptance of service robots’ anthropomorphic features.Online questionnaire is designed and distributed,then it’s reliability and validity is calculated.Hypothesises are tested through multiple regression and mediating effect analysis.The results show that the anthropomorphic features have a significant positive effect on user acceptance through psychological perception and emotional experience.(2)Analyzing physiological mechanism of effect on user acceptance based on a multimodal measurement method.Anthropomorphic visual features of service robots are evaluated and representative service robot images are selected.Multimodal measurement experiment is designed and conducted.Subjective emotional experience indexes,eye movement indexes and physiological indexes are chosen from raw data.The influence of anthropomorphic visual features on the users’ acceptance and emotional indexes are analyzed as well as correlation between acceptance and emotional indexes.The results show that influence on acceptance is ranked from high to low as anthropomorphism at modereate level,anthropomorphism at high level and anthropomorphism at low level.The subjective emotional experience indexes with significant differences are pleasure and arousal;the eye movement indexes with significant differences are pupil diameter,fixation counts,saccade counts and saccade velocity;the physiological indexes with significant differences are RMS,MPF,SCL and Mean HR.The results of paired sample t-test show different effects of diverse anthropomorphic visual features on emotional indexes.In terms of correlation,there is a significantly positive correlation separately between pleasure and acceptance,arousal and acceptance separately;there is a significantly positive correlation separately between pupil diameter and acceptance,saccade counts and acceptance,saccade velocity and acceptance,but no significant correlation between fixation counts and acceptance;there is a significantly negative correlation between MPF and acceptance and a significantly positive correlation separately between RMS and acceptance,SCL and acceptance,Mean HR and acceptance.(3)Analyzing brain mechanism of effect on user acceptance based on an ERP measurement method.An ERP experiment is designed and conducted based on amended Oddball paradigm.Subjective emotional experience indexes,P2,P3 and LPP component are chosen from raw data.The influence of anthropomorphic visual features on user acceptance、subjective emotional experience indexes and ERP components are analyzed as well as correlation between acceptance and emotional indexes The results show that anthropomorphic visual features at moderate level induced significant larger P2 amplitude than that induced by features at high and low level in 160 ～ 200 ms after stimulation in central area.Anthropomorphic visual features at moderate level induced significant larger P3 amplitude than that induced by features at high and low level in400 ～ 600 ms after stimulation in parietal area.Anthropomorphic visual features at moderate level induced significant larger LPP amplitude than that induced by features at high and low level in 600 ～ 800 ms after stimulation in central-parietal and parietal area.(4)Proposing acceptance prediction model based on multimodal data.Emotion measurement indexes are selected as the input parameters of machine learning and neural network prediction model,prediction accuracy of different indexes combinations is compared and the optimal combination is found.The results show that the comprehensive use of subjective emotional experience indexes(pleasure and arousal),eye movement indexes(pupil diameter,saccade counts and saccade velocity)and physiological indexes(RMS,MPF,SCL and Mean HR)has prediction accuracy of 99.33% in BPNN.