Electrocardiogram (ECG) signal processing for emotion recognition has become a promising area, particularly for applications in personalized experiences, human–computer interface (HCI), and mental health. ECG-based techniques are less susceptible to manipulation and offer a trustworthy measurement through physiological indications like heart rate variability (HRV), in contrast to traditional emotion identification techniques that rely on visual or auditory cues. This paper explores the emotion recognition applying deep learning models using the ECG signals. The model accuracy is enhanced by integrating the convolutional neural network (CNN) and long short-term memory (LSTM) models and incorporating the temporal and spatial data. The hybrid model resulted in a high accuracy of 92.11% for arousal and 90.79% for valence, on the DREAMER dataset. This enhancement in emotion recognition using the ECG signals has a great impact across a wide range of applications.