This study aims to develop a low-cost and simple algorithm for detecting sleep apnea at home using only continuous audio data. We propose a real-time monitoring model consisting of a combination of long-short term memory (LSTM) and convolutional neural networks (CNNs). The model can continuously receive audio data from the subject and predict whether the subject has exhibited sleep apnea symptoms during the corresponding period. By utilizing the continuous input and output capabilities of LSTM, along with the image recognition features of CNNs, we have created an optimal algorithm for continuous monitoring. This method allows subjects to perform real-time detection in daily life, providing information that is more closely aligned with their normal sleep patterns. Compared to the complicated detection procedures in hospitals, this method can quickly, massively, and immediately screen for suspected sleep apnea patients, making it more universal.