Parkinson’s disease (PD) is a neurodegenerative disorder caused by a lack of dopamine secretion, resulting in motor dysfunction that affects speech production. The literature indicates that PD can be identified by observing changes in speech signals over time. Early detection of PD is essential to slow its progression and enable patients to access disease-modifying therapies. In this context, an adapted Convolutional Neural Network (CNN) for this task appears promising. In this paper, we propose a CNN classifier to efficiently detect PD with an accuracy above 95\%. We employed advanced machine learning techniques to identify PD early through voice analysis using the Parkinson’s Disease Voice Recording Data Set, which includes the vocal characteristics of 147 PD patients and 48 healthy individuals. The primary objective is to enhance the diagnostic accuracy of PD using a CNN tailored for this task. The preprocessing steps involved cleaning and normalizing the data, followed by the application of Bayesian hyperparameter optimization to improve model performance. The CNN model achieved an average accuracy of 96.8451\% in PD detection, demonstrating high precision, sensitivity, and a balanced F1 score. This result underscores the model’s effectiveness in identifying individuals with PD and highlights the potential of using a single CNN topology, as opposed to more complex hybrid models, to achieve satisfactory diagnostic rates.To evaluate the performance of the proposed approach, we used simulated data generated from the original dataset. The results indicate that the proposed approach is suitable for the automatic identification of PD in practical scenarios.