Polysomnography (PSG)-based sleep stage scoring is time-consuming and it suffers from variability in results. With the help of automated PSG scoring, which is based on deep learning techniques, it is possible to reduce labor costs and variability inherent to this task. Instead of using publicly available sleep databases, we created our database by using the Philips Alice clinic device, which employs 19 sensor channels connected to subjects. Since every sensor channel creates a huge amount of data, we limited our work to 50 patients and pre-processed this data. The deep learning sleep-based sleep stage classifier demonstrates excellent accuracy and agreement with the sleep expert's scoring. Average accuracy, precision, recall, and F1-measure were defined as 91.6, 90.9, 91.6, and 90.7% respectively. The proposed work has novelty when it is compared with similar deep learning-based automatic sleep staging studies by using 19 channels as input rather than employing only EEG, EOG, or EMG data. The proposed model is scientifically created for making job quite similar to the sleep doctors during sleep stage scoring automatically instead of manual steps used by them. Furthermore, as explained, our database is created by using a local hospital, which has a sleep clinic, rather than using publicly available databases. The proposed work will allow a fully automated PSG scoring system by having its own database and employing all PSG inputs.