Research On Online Detection Of Freezing Of Gait For Parkinson’s Disease Patients Based On Wearable Device

Approximately half of Parkinson’s patients have suffered from freezing of gait,which often causes the patient to fall,interferes with daily activities and severely impairs the quality of life.On this basis,an online detection system for the freezing of gait of Parkinson’s patients using wearable devices is proposed.The system can detect the gait data of Parkinson’s patients in real time,and pass a rhythm when the patient suffered from freezing of gait.Sexual auditory stimulation to achieve the effect of intervention on freezing of gait.The main research work is as follows:(1)An online detection system for frozen gait of Parkinson’s patients based on wearable devices was constructed,in which the data collection equipment was placed on the patient’s waist,left and right thighs,and left and right calves;Secondly,the experimental data is studied by sliding window to obtain positive and negative sample sets.(2)The sample set is extracted from the four aspects of time domain、frequency domain、FI index and other related features.There are 54 features that can be extracted on each axis of the sensor.Aiming at problems such as over-fitting caused by multi-dimensional features,the top 35 features were selected through preprocessing,non-parametric testing,and maximum information coefficients.(3)In-depth research has been carried out on the related factors that affect the performance of the algorithm,such as the sensor configuration combination,the ratio of positive and negative samples,the window length and step length of the divided samples.The detection effects of six different sensor configuration combinations,four positive and negative sample ratios,eight different window length and step length combinations,and six different feature numbers are put into model training.The results are displayed in the window length 3s,step length In the case of 0.5s and a 2:1 positive-negative sample ratio,the top 25 features are trained in the logistic regression algorithm to obtain the best detection effect.The sensitivity is 91.36% and the specificity is 87.08%.The detection system performs online detection of freezing of gait,and proposes four new detection indicators based on freezing of gait events.The results show that the average hit rate of 8 patients is 92.4%,and the false alarm rate is0.83Times/min,the average delay time is 0.25 s,and the average early hit time is 1.5s.