Parkinson's Disease is a debilitating neurodegenerative disorder that affects movement and coordination. Early detection of the disease is critical for proper treatment, and machine learning techniques can provide a non-invasive and effective way of predicting Parkinson's Disease from speech signals. In this paper, a novel deep-learning model called DLN-PD is proposed to detect Parkinson's Disease from speech signals. The dataset from UCI is used to train and evaluate the model. The proposed model is empirically compared with three other models namely KNN-PD, LSTM-PD, ANN-PD, and CNN-PD which are based on KNN technique, LSTM model, ANN technique, and CNN model respectively. The experimental results show that the proposed model DLN-PD has a significant potential for Parkinson’s disease detection over voice signals with an accuracy of 97.6% and an AUC of 98.2%.