Obstructive sleep apnea and hypopnea syndrome (OSAHS) causes excessive daytime sleepiness and cognitive decline due to prolonged nocturnal hypoxia. Without timely treatment, it increases the risk of obesity, ischemic heart disease, stroke and other serious disorders. The standard detection method, nocturnal polysomnography (PSG), is expensive and available only in limited facilities. The study proposed a cheaper approach using electrocardiogram (ECG) signals and biometric data. The method integrates ECG features extracted by a long-short-term memory (LSTM) network with biometric features obtained by support vector machines (SVM), classified by a fully connected layer. The LSTM-SVM model achieved an accuracy of 97.1% on two independent databases and 92% on a separate dataset. The results demonstrate strong generalizability and potential for clinical application.