Emotional stress often has a significant effect on the working performance of staff, but this effect is commonly neglected in existing staff scheduling methods. We study a call-center staff scheduling problem, which considers the evolution of work performance of staff under emotional stress. First, we present an emotional stress driven model that estimates the working performance of call-center employees based on not only skill levels but also emotional states. On the basis of the model, we formulate a combined short-term and long-term call-center staff scheduling problem aiming at maximizing the customer service level, which depends on the working performance of employees. We then propose a memetic optimization algorithm combining global mutation and neighborhood search assisted by deep reinforcement learning to efficiently solve this problem. Experimental results on real-world problem instances of bank call-center staff scheduling demonstrate the performance advantages of the proposed method over selected popular staff scheduling methods. By explicitly modeling and incorporating emotional stress, our method reflects a more realistic understanding and utilization of human behavior in staff scheduling.