【作者】 李明；

【导师】 张秋豪； 张冠锋；

【作者基本信息】 沈阳工业大学 ， 电气工程， 2025， 硕士

【摘要】 随着我国老龄化进程加剧,养老问题日益凸显,已成为我国急需解决的重要问题之一,尤其面临护工短缺的困境。近年来电动轮椅的智能化在一定程度上解决了部分护工资源不足的压力。然而,目前的智能轮椅大都针对自主避障、移动控制等方面进行优化,往往未能充分考虑服务过程中兼顾用户情感的问题,服务效果未能达到最佳。本研究构建了一个整合情绪识别、主观评估与生理反应的动态情绪评价体系,突破传统模型过度依赖客观指标的局限性。以Russell二维情绪模型为理论根基,结合SAM自我评估量表设计标准化评估流程,通过系统性训练强化受试者情绪数字化表达能力。在此基础上,将智能轮椅控制系统与虚拟现实技术融合,进行深度情绪诱发,搭建具有独立变量控制能力的沉浸式实验场景,并设计基于智能轮椅不同移动过程的情绪诱发实验,同步采集受试者在实验中的脑电信号、面部表情系数及情绪主观评价。为研究智能轮椅不同移动过程对用户情绪影响、建立情绪识别模型提供数据支持。本研究针对智能轮椅不同移动过程对用户情绪的影响分别对实验数据从情绪主观评价与情绪生理反应层面进行了深入探讨,其中,对于情绪主观评价进行统计分析,对于情绪生理反应进行脑电地形图分析,最后总结出智能轮椅不同移动过程对用户情绪的影响与用户情绪的生成机制。本研究运用CNN—LSTM框架分别针对脑电信号、面部表情以及两种模态融合构建了三种情绪识别网络,旨在从不同信号输入的角度实现智能轮椅不同移动过程下用户情绪识别,三种情绪识别网络准确率分别为80.56%、70.83%、84.72%,随后使用Deap数据集进一步验证模型性能,分析结果显示,训练后的模型能够准确识别用户在智能轮椅移动过程中的情绪状态且具有一定泛化能力,为智能轮椅识别用户情绪提供了有效方法。更多还原

【Abstract】 With the accelerating aging process in China,the issue of elderly care has become increasingly prominent and is now one of the critical challenges requiring urgent resolution,particularly the shortage of caregivers.In recent years,the intelligentization of electric wheelchairs has alleviated some pressure on the insufficient caregiver resources to a certain extent.However,existing intelligent wheelchairs primarily focus on optimizing autonomous obstacle avoidance and mobility control,often neglecting the emotional needs of users during service interactions,resulting in suboptimal service outcomes.This study constructs a dynamic emotion evaluation system that integrates emotion recognition,subjective evaluation and physiological response,breaking through the limitations of traditional models that rely too much on objective indicators.Based on Russell’s two-dimensional emotional model,combined with SAM self-assessment scale,a standardized assessment process was designed to strengthen the digital expression ability of subjects’emotions through systematic training.On this basis,the intelligent wheelchair control system is integrated with virtual reality technology to carry out deep emotion induction.An immersive experimental scene with independent variable control ability is built,and an emotion induction experiment based on intelligent wheelchairs different movement processes is designed.The EEG signals,facial expression coefficients and subjective emotional evaluation of the subjects in the experiment are collected synchronously.It provides data support for studying the influence of different movement processes of intelligent wheelchairs on users’emotions and establishing an emotion recognition model.This study focuses on the influence of different mobile processes of intelligent wheelchairs on user emotions.The experimental data are discussed in depth from the aspects of emotional subjective evaluation and emotional physiological response.Among them,statistical analysis is carried out for emotional subjective evaluation,and EEG topographic map analysis is carried out for emotional physiological response.Finally,the influence of different mobile processes of intelligent wheelchairs on user emotions and the generation mechanism of user emotions are summarized.In this study,the CNN-LSTM framework was used to construct three emotion recognition networks for EEG signals,facial expressions,and two modal fusions,aiming to realize user emotion recognition under different mobile processes of intelligent wheelchairs from the perspective of different signal inputs.The accuracy of the three emotion recognition networks was 80.56%,70.83%,and 84.72%,respectively.Then the Deap data set was used to further verify the performance of the model.The analysis results show that the trained model can accurately identify the user’s emotional state during the movement of the intelligent wheelchair and has a certain generalization ability,which provides an effective method for intelligent wheelchairs to identify user emotions.更多还原