Abstract
Abstract
In the training and competition process of athletes, their bodies are subjected to various levels of load and stress. As an important diagnostic tool, ECG signals can provide deep insights into the cardiac function of athletes, including heart rate, rhythm, and changes in cardiac electrical activity. By conducting a thorough examination of ECG readings, we are able to quickly identify possible heart conditions or irregularities, which is essential for preserving the heart health of athletes. However, ECG signals are highly complex and multidimensional. To accurately classify these signals, it is necessary to select the most representative and discriminative features. However, this is not an easy task, and the selection of effective features remains a pressing issue. To address this problem, this paper proposes the CSNet classification network model. This framework eradicates disruptions in electrocardiogram signals, performs attribute extraction via a direct network configuration, and combines channel focus mechanisms and spatial focus mechanisms to enhance attribute representation and categorization capabilities. Furthermore, to retain the temporal information of ECG signals, we introduce the Gated Recurrent Unit (GRU), which helps to better capture temporal patterns and dependencies in the signals, thus enabling more accurate classification of ECG signals.