The Development Of Parkinson’s Disease Detection System Based On Speech

According to the World Parkinson Disease Association statistics, there were already more than 500 million patients with Parkinson disease. It has become the third largest disease in the field of the nervous field, ranked only second to stroke and epilepsy. However, when perform Parkinson Diagnosis, The current method of detection is based on the long-term clinical observation performed by the doctors who can draw conclusions from the tester’s behavior. Obviously, this method over-reliance on the doctors’ ability, and it need a long detection period as well as high cost. Numerous studies show that there is a link between Parkinson’s disease and dysphonia. Besides, one of the Parkinson`s symptoms is the low volume, pronounced tremor. Apparently, we can use speech to diagnose this diesase. Speech signals fit ideally the purpose of telemonitoring, because they are non-invasive, can be self-recorded, and are easy to obtain from a subject who is not expected to perform any special kinds of actions in order to record his voice. For those reasons, using speech to detect Parkinson has attracted more and more attention.Considering the above background, this paper studied Parkinson detection method based on speech and developed a prototype system. The system includes three models, they are Voice collecting, speech feature extraction and detection. The Voice acquisition module is used to record voice and display final results; Speech feature extraction module is used to extract speech pathological features; in the third module, we use support vector machine technology to classify the Pathological features. To speed up the extraction of speech features, this paper also parallelizes pitch detection algorithm.This system has solved shortcomings of the traditional method such as Time-consuming, labor-intensive, high cost. In the current "Internet +" background and trend of population aging, this system has a great market prospects.