| Our society is facing with the issue of the population aging,and it is accompanied by the movement inconvenience of the elderly population.For this,the research on flexible and safe assistant robot is very important to solve the problem of inconvenience of the elderly.Although the compliant walking assistance method based on the user's intention has been extensively studied,the transfer behavior from one state to another such as getting up,standing,and the like has not been sufficiently studied.The multi-transfer behavior will have a greater impact on the user's sole,lower back and legs.Therefore,our laboratory independently developed several welfare robots,including intelligent wheelchair,smart bed,auxiliary standing robot,etc.,to achieve the basic behavioral assistance and transfer behavior assistance of daily life.At the same time,the researches about the fatigue of users are not abundant.Especially for patients with dyskinesia,they need a large mental workload and physical load when completing a task.Therefore,the research on mental workload for multi-transfer behavior is particularly important.Electromyogram(EMG)signals and electroencephalography(EEG)signals are usually used to evaluating users' fatigue during the process of human-computer interaction.These methods cannot accurately measure the users' mental workload,however cerebral blood oxygen signals can accurately display brain activity.Now they are used for analysis of the mental workload of pilots.Therefore,this thesis proposes using cerebral blood oxygen signal to evaluate mental workload during the assistant transfer process.Firstly,this thesis uses near-infrared spectroscopy to achieve non-destructive detection of cerebral blood oxygen signals.In order to realize the collection of cerebral blood oxygen signals during the similar transfer behavior,an easy-to-carry onboard motion attitude monitoring system is designed to achieve the attitude recognition and time calibration while the transfer processes are realized.At the same time,Gaussian filtering algorithm is used to smooth the waveform to achieve the purpose of removing noise.Secondly,this thesis builds a combined mental workload assessment model by proposing an optimal feature parameter combination classification method through time domain analysis,frequency domain analysis and approximate entropy analysis.Through the support vector machine and linear discriminant analysis algorithm,the recognition rates of83.3% and 80.5% are obtained respectively,and it is concluded that SVM(Support Vector Machine)algorithm with the mean value and variance as the eigenvalue input has the highest recognition rate.Then,this study extracted the mean and variance eigenvalues of the blood oxygen of(Dorsolateral prefrontal lobe,DLPFC)of subjects when they performed 3-level n-back task,Then mean and variance eigenvalues are inputted into the SVM algorithm for classification.The output is the three levels of mental workload,labeled as level 1,level 2 and level 3.And,the cerebral blood oxygen content corresponds to the activation degree of the corresponding brain regions.If some exercise postures cause the blood oxygen content in the mental workload sensitive areas to be higher than the resting value,which indicates that the action causes an increase in mental workload.Combining the mental workload level with the response of mental workload-sensitive area can more accurately distinguish the mental load in different states.Finally,this thesis designs the experiments to study the motion posture affecting the mental load during the transfer process,and finds that when the bed is used to assist the movement,the amplitude and dimension of the subject's motion posture become smaller and the activation degree of the mental workload sensitive area becomes lower.It is concluded that the high-dimensional large-scale movement of the limb increases the mental workload with a high probability. |
PDF Full Text Request