HUMAN SOUND BASE DISEASE DETECTING SYSTEM USING RESNET 18

Recent developments in ML and AI have opened up new possibilities for the identification and diagnosis of diseases. Here, we provide a new approach to human illness detection using deep convolutional neural networks (CNNs) based on the ResNet-18 architecture. By analyzing the patterns of sounds made by people when they speak or cough, the system hopes to reliably diagnose a range of illnesses. The suggested technique makes use of deep learning to efficiently classify diseases by automatically extracting important characteristics from audio inputs. Spectrograms and other representations of human sound data may be processed using ResNet-18, a network famous for its success in image recognition tasks. This network is able to capture subtle patterns that indicate various disorders. Gathering data, cleaning it up, extracting features, training the model, and finally classifying diseases are all crucial parts of the system. For both training and validation, we use large datasets that comprise audio recordings from people with various medical problems. Using transfer learning methods, the ResNet-18 model is trained to take advantage of pre-learned weights, which speeds up convergence and improves performance. In order to determine how well the system works, we do lengthy trials with various datasets that include a broad variety of ailments, such as respiratory issues, heart problems, and neurological impairments. When it comes to detecting different illnesses, the algorithm shows promise with high recall rates and accuracy. In addition, the suggested system has a number of benefits, such as being non- invasive, affordable, and scalable. The system may be simply implemented in many healthcare settings by using common sound recording devices like cell phones or specialized sensors. This allows for the early diagnosis and intervention of diseases.