Research And Implementation Of AD Auxiliary Diagnosis Model Based On Acoustic And Linguistic Features

Alzheimer’s disease is a neurodegenerative disease that has a higher incidence and greater harm in middle-aged and elderly people.At present,no specific drugs for Alzheimer’s disease have been developed.In medical practice,only corresponding therapeutic interventions can be taken according to the stage of the patient’s disease to delay the development of the disease.Therefore,it is necessary to determine the stage of the patient’s disease as soon as possible.treatment is important.In contrast,using a patient’s voice to assist in the diagnosis of Alzheimer’s disease has proven to be a Simple and efficient new diagnostic method.This paper takes the speech samples of patients with Alzheimer’s disease as the research object,and proposes an AD auxiliary diagnosis model based on the acoustic features and language features of the subjects’ speech.According to the definition in medicine,the model divides the subjects’ speech into three categories: healthy control（HC）,mild cognitive impairment（MCI）and Alzheimer’s disease diagnosis（AD）.After data preprocessing,feature extraction,and model training,the patient’s disease stage can be classified.1)According to the acoustic features extracted from the subjects’ speech samples,this paper proposes an AD auxiliary diagnosis model based on the acoustic features of the subjects’ speech.The model is built based on the Alex Net network,and on this basis,the CBAM attention mechanism is introduced to reduce the influence of the interference features contained in the speech on the classification performance of the model.The experimental results show that the model proposed in this paper can effectively use the acoustic features in the speech to complete the classification of the subjects’ disease stages,and the classification accuracy rate reaches 84.8%.2)Taking the linguistic features in the speech samples as the research object,this paper proposes an AD auxiliary diagnosis model based on the linguistic features of the subjects’ speech.The model is based on the pre-trained language model DC-BERT.The utterance-level feature extraction ability of BERT and the feature classification ability of the attention-enhanced neural network model proposed in this paper are combined to realize the classification of disease stages based on the linguistic features of the subjects’ speech.3)Based on the proposed AD auxiliary diagnosis model,this paper designs and implements an AD voice auxiliary diagnosis system,which is published on the We Chat applet platform.