According to the American National Heart, Lung, and Blood Institute (NHLBI), undiagnosed Sleep Apnea (SA) may raise the risk of high blood pressure, diabetes, heart disease, and stroke. A lot of these complications can be managed effectively or even prevented through early diagnosis of SA, which often goes undetected since it occurs mostly during the natural daily sleep cycle. Thus a low cost and non-invasive solution is urgently needed that may be used for home monitoring over longer periods, instead of high cost and obtrusive clinical methods such as polysomnography. Towards this goal, we propose a simple yet effective computational technique that combines feature fusion with machine learning (ML) in an interpretable manner from electrocardiogram (ECG) signals with high fidelity demonstrated through strong experimental results on benchmark public datasets. We have deliberately chosen a non deep learning based ML method that is transparent and fast, because this work is meant to be a proof of concept for future translation into practical application using smart portable ECG monitors, thereby facilitating early detection of SA and improve patient health.