Parkinson’s disease (PD) is a neurodegenerative disorder that causes gait abnormalities. Early and accurate recognition of PD gait is crucial for effective treatment. Recently, deep learning techniques have shown promising results in PD gait analysis. However, most existing methods focus on severity estimation and frozen gait detection, while the recognition of Parkinsonian gait and normal gait from the forward video has not been reported. In this paper, we propose a novel spatiotemporal modeling method for PD gait recognition, named WM–STGCN, which utilizes a Weighted adjacency matrix with virtual connection and Multi-scale temporal convolution in a Spatiotemporal Graph Convolution Network. The weighted matrix enables different intensities to be assigned to different spatial features, including virtual connections, while the multi-scale temporal convolution helps to effectively capture the temporal features at different scales. Moreover, we employ various approaches to augment skeleton data. Experimental results show that our proposed method achieved the best accuracy of 87.1% and an F1 score of 92.85%, outperforming Long short-term memory (LSTM), K-nearest neighbors (KNN), Decision tree, AdaBoost, and ST–GCN models. Our proposed WM–STGCN provides an effective spatiotemporal modeling method for PD gait recognition that outperforms existing methods. It has the potential for clinical application in PD diagnosis and treatment.

Hand tremor is a major motor symptom of Parkinson’s disease (PD) that seriously affects the life quality of patients. Conventional tremor detection methods have mainly focused on resting state. However, resting tremor is not always sufficient to accurately distinguish PD tremor from other tremor disorders such as essential tremor. Examining tremor during movement provides additional diagnostic value, but also has challenges. In this paper, we propose a novel approach to detect Parkinsonian hand tremor at walking state by exploiting the periodic synchronization between arm swings and leg steps that is observed in natural walking, specifically by using a wrist-worn sensor to capture the characteristic hand tremor that is induced during walking in Parkinson’s patients. Our method first applies a dynamic scanning mechanism to screen out valid walking fragments from whole walking sequence. We then exploit walking periodicity to combine time and frequency domain features, and design two new accelerometer-based features, Average Peak Interval Time (APIT) and Average Peak Frequency (APF), which capture the characteristic amplitude and frequency patterns of PD tremors during natural walking movements. Finally, unlike traditional approaches that use uniform feature weighting, our proposed variance adaptive AdaBoost (VA-AdaBoost) algorithm uniquely incorporates accelerometer signal variance to adaptively weight samples for feature training. The results of our five-fold cross-validation experimental evaluation using the PD-BioStampRC21 dataset show that our method achieved an accuracy of 86.67% and an AUC score of 95.56%, representing a significant improvement over the baseline model’s accuracy of 72.67% and surpassing other existing methods. Our method enables more accurate detection of subtle tremor patterns in real-world conditions compared to conventional methods. The proposed method has shown potential in the integration of wearable sensors and ensemble learning techniques for the early and accessible diagnosis of Parkinsonian hand tremor, ultimately contributing to improved patient care and treatment.