Using Deep Learning for Task and Tremor Type Classification in People with Parkinson's Disease.

Hand tremor is one of the dominating symptoms of Parkinson's disease (PD), which significantly limits activities of daily living. Along with medications, wearable devices have been proposed to suppress tremor. However, suppressing tremor without interfering with voluntary motion remains challenging and improvements are needed. The main goal of this work was to design algorithms for the automatic identification of the tremor type and voluntary motions, using only surface electromyography (sEMG) data. Towards this goal, a bidirectional long short-term memory (BiLSTM) algorithm was implemented that uses sEMG data to identify the motion and tremor type of people living with PD when performing a task. Moreover, in order to automate the training process, hyperparamter selection was performed using a regularized evolutionary algorithm. The results show that the accuracy of task classification among 15 people living with PD was 84±8%, and the accuracy of tremor classification was 88±5%. Both models performed significantly above chance levels (20% and 33% for task and tremor classification, respectively). Thus, it was concluded that the trained models, based on using purely sEMG signals, could successfully identify the task and tremor types.

Implementing Machine Learning Algorithms to Classify Postures and Forecast Motions When Using a Dynamic Chair.

Many modern jobs require long periods of sitting on a chair that may result in serious health complications. Dynamic chairs are proposed as alternatives to the traditional sitting chairs; however, previous studies have suggested that most users are not aware of their postures and do not take advantage of the increased range of motion offered by the dynamic chairs. Building a system that identifies users' postures in real time, as well as forecasts the next few postures, can bring awareness to the sitting behavior of each user. In this study, machine learning algorithms have been implemented to automatically classify users' postures and forecast their next motions. The random forest, gradient decision tree, and support vector machine algorithms were used to classify postures. The evaluation of the trained classifiers indicated that they could successfully identify users' postures with an accuracy above 90%. The algorithm can provide users with an accurate report of their sitting habits. A 1D-convolutional-LSTM network has also been implemented to forecast users' future postures based on their previous motions, the model can forecast a user's motions with high accuracy (97%). The ability of the algorithm to forecast future postures could be used to suggest alternative postures as needed.

Improved ozone DIAL retrievals in the upper troposphere and lower stratosphere using an optimal estimation method.

We have implemented a first-principle optimal estimation method to retrieve ozone density profiles using simultaneously tropospheric and stratospheric differential absorption lidar (DIAL) measurements. Our retrieval extends from 2.5 km to about 42 km in altitude, and in the upper troposphere and the lower stratosphere (UTLS) it shows a significant improvement in the overlapping region, where the optimal estimation method (OEM) can retrieve a single ozone profile consistent with the measurements from both lidars. Here stratospheric and tropospheric measurements from the Observatoire de Haute Provence are used, and the OEM retrievals in the UTLS region compared with coincident ozonesonde measurements. The retrieved ozone profiles have a small statistical uncertainty in the UTLS region relative to individual determinations of ozone from each lidar, and the maximum statistical uncertainty does not exceed a maximum of 7%.