论文部分内容阅读
基于语音数据实现帕金森病诊断近年来已被证明是一种有效方式。但是,目前相关研究在样本预处理和集成学习方面还考虑不足,从而造成样本对分类器误导、分类准确率和稳定性还不令人满意等问题。本文提出了一种结合样本重复剪辑算法和随机森林的帕金森病诊断新算法,并基于最新公共数据集进行了对比实验。实验结果表明,本文算法实现了对语音样本和受试者的分类诊断,针对受试者的平均分类准确率达到了100%,比原数据提供者最高改善了29.44%。本文基于样本优选实现了一种新的语音帕金森病诊断算法;与同类算法相比,具有较高的准确率和稳定性。
Diagnosis of Parkinson’s disease based on voice data has proven to be an effective way in recent years. However, the current researches are still insufficient in sample preparation and integrated learning, which leads to the sample misclassification of classifiers, the accuracy and stability of classification are not satisfactory. In this paper, we propose a new algorithm for the diagnosis of Parkinson’s disease combined with sample-repeat clipping algorithm and random forest, and make comparative experiments based on the latest public data set. The experimental results show that the proposed algorithm achieves classification diagnosis of speech samples and subjects with an average classification accuracy of 100% for subjects and a maximum improvement of 29.44% over original data providers. In this paper, a new speech diagnosis algorithm for Parkinson’s disease based on sample optimization is achieved. Compared with similar algorithms, it has higher accuracy and stability.