This paper provides a comprehensive review of machine learning (ML) techniques for automated human emotion recognition (AHER). It highlights the interdisciplinary nature of emotion research, which involves psychology, philosophy, and computing. The review covers various modalities used in AHER, including facial expressions, written text, speech, and physiological signals. The methods can beuni-modal or multi-modal, and recent research focuses on in-the-wild experiments and generic models. The paper discusses the evolution of emotion research, from Darwin's early theories to modern emotion models, and categorizes these models into categorical and dimensional types. It also delves into the application of ML algorithms, such as classification and regression, in emotion recognition, detailing both shallow and deep learning techniques. The review emphasizes the importance of supervised and unsupervised learning methods, including linear and nonlinear classification algorithms, regression methods, and deep learning models like deep belief networks, convolutional neural networks, and generative adversarial networks (GANs). The paper concludes with a discussion on the strengths and weaknesses of the reviewed studies and provides recommendations for future research directions.This paper provides a comprehensive review of machine learning (ML) techniques for automated human emotion recognition (AHER). It highlights the interdisciplinary nature of emotion research, which involves psychology, philosophy, and computing. The review covers various modalities used in AHER, including facial expressions, written text, speech, and physiological signals. The methods can beuni-modal or multi-modal, and recent research focuses on in-the-wild experiments and generic models. The paper discusses the evolution of emotion research, from Darwin's early theories to modern emotion models, and categorizes these models into categorical and dimensional types. It also delves into the application of ML algorithms, such as classification and regression, in emotion recognition, detailing both shallow and deep learning techniques. The review emphasizes the importance of supervised and unsupervised learning methods, including linear and nonlinear classification algorithms, regression methods, and deep learning models like deep belief networks, convolutional neural networks, and generative adversarial networks (GANs). The paper concludes with a discussion on the strengths and weaknesses of the reviewed studies and provides recommendations for future research directions.