The breathing signal is used to check the position possibility of disorders that can cause chronic diseases and dysfunctions eases that occur during sleep, such as obstructive sleep apnea (OSA). Breath acquisition equipment is often uncomfortable for patients and produces uncertain results. An alternative to these problems were predicting breathing using electrocardiogram signs (ECG) and photoplethysmography (PPG). Those physiological signals are related to respiratory activity, so the idea of the work is to use Deep Learning techniques with target audience specific dataset of these signals to estimate the respiratory signal in an intra-subject manner, where each of the patients went through the training model in the topology shown in this study. The successful technique was the Convolutional Neural Network (CNN), which obtained an average mean squared error (MSE), Mean Absolute Error (MAE) and Coefficient of determination (R-squared) with the record of 10 patients.