| Cognitive workload refers to the amount of mental resources or the proportion of cognitive resources that people need to provide when performing related tasks.Excessive cognitive workload could lead to a decline in execution efficiency and an increase in error rate,which could easily lead to accidents in high-risk industries;and suffering long-term high cognitive workload will also hinder the development of people’s physical and mental health.Therefore,it is essential to monitor cognitive workload level.At present,there are a variety of physiological detection methods for cognitive workload,such as electroencephalography(EEG),electrocardiogram(ECG).But most of these detection methods cause a sense of discomfort or need a long detection period.Pulse wave contains abundant physiological information of the human body,and the signal is easy to collect and the equipment is portable.It is suitable for wearable devices to realize all-weather monitoring of cognitive workload.The method of detecting cognitive workload with pulse wave is mainly based on morphological features and pulse rate variability features.In this thesis,a cognitive workload detection method based on pulse components’ features is proposed to reduce the difficulty of morphological features extraction and shorten the detection time.The main work includes the following aspects:(1)Building a pulse wave signal acquisition system to collect the pulse wave signals of participants in the cognitive workload experiment.The design content of the pulse wave signal acquisition system includes three parts: infrared transceiver analog circuit design,FPGA-based digital acquisition and control system design,and real-time interactive interface development.The cognitive workload experiment is based on the dual-stimulus N-Back experimental paradigm,with multiple difficulty levels set to induce subjects to produce different levels of cognitive workload.(2)In signal processing and feature extraction,a complete set of pulse wave signal processing and pulse component extraction scheme is presented in this thesis.Including the maximum slope point reverse traversal algorithm to enhance the robustness of the base point location,the pulse wave self-extraction template to detect motion artifact noise,and the Gaussian pulse component extraction algorithm.The pulse components of the pulse wave signal are extracted through the above scheme,and a set of pulse component features are obtained by calculation in the time domain and frequency domain,which are used as a data set for cognitive workload detection and modeling.(3)Based on machine learning algorithms,the effectiveness of pulse component features in cognitive workload detection is verified.The features’ change and distribution differences among different cognitive workload states were analyzed with statistical tools.w The performance of the pulse components’ features under the detection window length of 1min was tested with various classifier algorithms,and the three-category detection accuracy of71.94% and the average two-category detection accuracy of 81.35% were obtained,respectively 20.38% and 15.68% higher than the detection method based on pulse rate variability characteristics.And it also has a higher temporal resolution.In the Subject-Independent Testing,when there is an obvious difference in cognitive workload level,the detection accuracy can be more than 70%,which means that the influence of individual differences on detection can be well overcome. |
