Pathology Classification in Voice Signal

Pathology Classification in Voice Signal

Kartikeya Hegde1, Karthik KS2, Karthik SR3

1Department of Computer Science And Engineering MITE Moodabidri 574225.

2Department of Computer Science And Engineering MITE Moodabidri 574225.

3Department of Computer Science And Engineering MITE Moodabidri 574225.

---------------------------------------------------------------------***---------------------------------------------------------------------

ABSTRACT - Voice pathology classification is the process of identifying different types of voice disorders from voice signals. With the advancement of machine learning techniques, researchers have explored the use of automated systems for the classification of different voice disorders. The process of pathology classification in voice signals involves several steps, including signal acquisition, feature extraction, and classification. The signal acquisition step involves recording the voice signal of the patient. The feature extraction step involves extracting relevant features from the voice signal. These features are then used to train a classification model using machine learning algorithms. From the voice datasets, create a Mel spectrogram for a particular voice from the dataset. The proposed approach employs a convolutional neural network (CNN) trained on a dataset of voices with three corresponding labels, which are Laryngol, normal, and vox senilis. The experiments show that MobileNetV2 achieves high accuracy compared to other models. The paper also discusses the performance of MobileNetV2, such as the number of layers and the learning rate. The accuracy of the pathology classification model depends on several factors, such as the quality of the voice signal, the choice of features, and the algorithm used for classification.

Key Words: Voice pathology, Signal acquisition, Mel spectrogram, three labels, MobileNetV2