This paper proposes that machine intelligence should include emotional intelligence, as it is crucial for human-like interaction. The authors demonstrate a method to recognize human affective states using four physiological signals. They collected data from a subject experiencing eight emotional states over multiple weeks, addressing challenges in obtaining reliable affective data. The study compares multiple algorithms for feature-based recognition of emotional states, analyzing four physiological signals with day-to-day variations. They propose new features and algorithms to handle these variations, achieving 81% recognition accuracy for eight emotional classes, including neutral. The paper highlights the importance of physiological emotion recognition in medicine, entertainment, and human-computer interaction. It discusses the challenges of gathering accurate physiological data, the importance of considering both internal and external expressions of emotion, and the need for reliable methods to elicit and measure emotional states. The authors also compare different techniques for feature extraction, selection, and transformation, including Sequential Floating Forward Search and Fisher Projection, to improve classification accuracy. The results show that combining these methods enhances performance, with the SFFS-FP approach achieving the highest accuracy. The study emphasizes the importance of physiological signals in understanding and recognizing emotions, and the need for further research in this area.This paper proposes that machine intelligence should include emotional intelligence, as it is crucial for human-like interaction. The authors demonstrate a method to recognize human affective states using four physiological signals. They collected data from a subject experiencing eight emotional states over multiple weeks, addressing challenges in obtaining reliable affective data. The study compares multiple algorithms for feature-based recognition of emotional states, analyzing four physiological signals with day-to-day variations. They propose new features and algorithms to handle these variations, achieving 81% recognition accuracy for eight emotional classes, including neutral. The paper highlights the importance of physiological emotion recognition in medicine, entertainment, and human-computer interaction. It discusses the challenges of gathering accurate physiological data, the importance of considering both internal and external expressions of emotion, and the need for reliable methods to elicit and measure emotional states. The authors also compare different techniques for feature extraction, selection, and transformation, including Sequential Floating Forward Search and Fisher Projection, to improve classification accuracy. The results show that combining these methods enhances performance, with the SFFS-FP approach achieving the highest accuracy. The study emphasizes the importance of physiological signals in understanding and recognizing emotions, and the need for further research in this area.