Parkinson's disease is a commonly observed neurological disorder that affects the nervous system and hinders, people's essential functions. The primary goal of this study is to identify the presence of Parkinson's disease by utilizing spectrogram images from voice recordings through the implementation of Convolutional Neural Networks (CNN). We conducted our research using a dataset from the Argentina. Our research made a significant contribution by performing various audio preprocessing operations. We split the audio samples into multiple segments of the same duration (2 seconds) and then implement audio augmentation techniques to increase the dataset. Finally, we converted these audio samples into spectrogram images to train our model. K-fold cross-validation method was used, set by (k=10) for further analysis. The model underwent 150 epochs of training, resulting in an Average Training Accuracy of 99.3% and an Average Testing Accuracy of 97.9%. The effectiveness of the proposed model is compared with five state-of-art models (AlexNet, VGG16, Inception V3, ResNet50, SqueezeNet) and the local binary pattern descriptors which were applied to the same dataset. As a result, the proposed model was found to be superior.