Optimize Deep Learning Model for Intensive Care of Neurological Disorders Patients Based on Facial Expression
Abstract
Facial expressions play a significant role in nonverbal communication. Reading the facial expressions of those suffering from neurological illnesses is crucial, as they may have significantly reduced verbal communication abilities. Such an assessment requires a thorough examination by medical specialists, which can be expensive and challenging. With the help of low-cost, non-invasive, automated facial expression detection technologies, experts can diagnose neurological disorders. In order to identify the facial expressions of individuals with Parkinson's, stroke, Alzheimer's, and Bell palsy disorders, this research constructs Optimized Deep Learning Model (FTDLM). The dataset is initially collected from well-known internet sites. Additionally, using publicly accessible sources, raw photos of the patient's most common facial expressions such as usual, happy, sad, and angry are gathered. Finding out if it was feasible to identify individual differences when searching for Parkinson's disease symptoms was the aim of the data analysis. Cropping is then taken into consideration in order to alter the image from the input image. Subsequently, the preprocessing method employing a Gaussian filter is examined to eliminate noise. Using FTDLM, the pre-processed image is used to classify the emotions. The New Convolutional Neural Network (NCNN) and the Enhanced Golden Search Algorithm (EGSA) are combined in this suggested model. EGSA is used in the NCNN to pick the hyperparameters. The suggested approach is carried out in Python, and statistical measures of accuracy, sensitivity, specificity, recall, and precision are used to assess performance. Furthermore, it is in contrast to traditional methods.