With the increase of daily human-computer-interactions and humans being highly emotional beings, came the research field of affective computing. More precisely, the need to incorporate emotions and emotion recognition into computer systems. Humans often express their emotions to other humans through body language, namely facial expressions, gestures and posture. Those can be deliberately influenced and to avoid this, physiological signals can be used for emotion recognition. They are regulated by the autonomous nervous system and influenced by emotions. This includes the electrical activity of the brain, heart and skin, as well as the skin temperature and respiration patterns, among others. Building systems that realize autonomous emotion recognition is a regularly studied topic and keeping an overview is difficult. An emotion recognition system usually consists of data acquisition, data processing and the assignment of emotion classes to the measured signals. Therefore, the foundation of emotion recognition systems includes the creation of data sets, the methods to extract and select features from the collected physiological signals and the classification algorithms. This thesis offers an explanation of the commonly used physiological signals, data sets and data processing methods, as well as an overview of some frequently used classifiers. However, the main objective is to provide an examination of topics that are currently being worked on. For this purpose, this thesis analyzes and reviews twenty-five studies published in the years 2018 to 2022 to report important findings or limitations and to highlight topics that are worth exploring further.