Emotion provides important details about human communication. During a conversation, it is customary to utilize facial expressions to express emotions. Furthermore, some interpersonal communication can be accomplished only through facial expressions. Some facial expressions are universal in that they convey the same feeling regardless of culture. If a machine could correctly perceive its user's facial expression, it might be able to assist them more quickly. This study creates a modular multi-channel deep convolutional neural network to improve face emotion recognition. A global average layer is used in the network output to avoid overfitting. The model's generalization ability can be improved by enhancing the dataset before training. Network offers a few advantages over other recognition algorithms. Finally, the trained recognition model is used to build a real-time face emotion recognition system. The results of the experiments will reveal that the system is capable of recognizing facial expressions in videos and photos. Keywords - convolutional neural network; network performance; expression recognition; machine learning; CNN; fer2013; TensorFlow;