Sleep apnea and hypopnea syndrome (SAHS) is a common yet potentially serious sleep disorder, affecting millions worldwide. It is characterized by pauses in breathing over 10 seconds long (sleep apnea) or as a shallow breathing during sleep (hypopnea). Those respiratory events disrupt the natural sleep cycle, leading to fragmented sleep and daytime fatigue. Higher risk of stroke, heart disease and high blood pressure, chronic fatigue, concentration and memory problems, depression and anxiety are just some examples of impacts on physical and mental health, caused by untreated SAHS. While being well documented and treatable (either by change of lifestyle or active therapy/surgery), most of the SAHS cases remain undiagnosed, either due to symptoms often being attributed to other factors or due to the fact, that the golden standard of SAHS diagnostics is polysomnography (PSG), which can be inaccessible for many. In the last couple of years, there have attempts to make SAHS detection more accessible, with detection from cardiac signals, such as electrocardiography (ECG) or photoplethysmography (PPG) having probably the biggest potential, being cheap and comfortable when compared to PSG, as both can be measured by wearable devices. In our work, we are aiming to utilize the bonds between cardiac signals and respiratory signals, by extracting multiple feature signals (changes in R-R or inter-beat intervals, fluctuations of amplitudes, signal drift), comparing different combinations of these signals and applying multiple deep learning methods (Alex-Net, LSTM and BiLSTM networks, U-net, hybrid networks) to binary (positive/negative) classify the SAHS, using Multi-Ethnic Study of Atherosclerosis database. All-night data from 150 patients with highest apnea-hypopnea index were used, giving approximately 10 000 healthy and 10 000 SAHS positive 1-minute intervals for training and testing, achieving accuracy up to 82.3% and sensitivity over 80.7%, giving possibility of millions SAHSs diagnosed early, without the need of PSG.