Maintaining mental health is crucial for emotional, psychological, and social well-being. Currently, however, societal mental health is at an all-time low. Robots have already proven useful in medicine, and robot assisted mental therapies through emotional monitoring have great potential. This paper reviews 60 recent papers to determine how accurately robots can classify human emotions using the latest sensor technologies. Among 18 different signals, it was determined that EDA sensors are best for this application. Our findings also show that CNN outperforms SVM, SVR, KNN and LDA for classifying EDA data with an average of 79% accuracy. This is further improved with the addition of RGB sensor data.