Automated Emotion Recognition (AER) is the process of programatically identifying and classifying affective responses to stimuli through the analysis of physiological signals. AER has applications in interpersonal communications via digital mediums, human-computer interactions, third-party monitoring and surveillance, personal health and wellness, and in physical and mental health treatment settings. Prior work largely relies on equipment that is most easily used in a laboratory environment. Wearable physiological sensors are now commonly found in smartwatches and fitness bands, opening the door to applications of AER in everyday life. In this paper, we demonstrate automated emotion recognition (AER) using deep learning with convolution neural networks (CNN) for automated feature extraction from ECG signals obtained using commercially available wearable ECG sensors. We utilize a novel approach to automated feature extraction relying on temporal CNN to resolve the time-dependent nature of the biomedical signal data. We achieve 96.2% accuracy in classifying emotional responses into the appropriate quadrant of the arousal / valence space.