This paper explores the development of machine emotional intelligence by focusing on the recognition of human affective states using physiological signals. The authors, Rosalind W. Picard, Elias Vyzas, and Jennifer Healey, argue that emotional intelligence is crucial for intelligent behavior and propose that machines should be capable of recognizing and responding to human emotions. They describe the challenges in obtaining reliable affective data and present a large dataset collected from a subject over multiple weeks, aiming to elicit and experience eight emotional states. The paper compares multiple algorithms for feature-based recognition of emotional states from this data, focusing on four physiological signals: electromyogram (EMG), photoplethysmography (PPG), skin conductance (SC), and respiration (R). The authors propose new features and algorithms to handle day-to-day variations in physiological signals, achieving an 81% recognition accuracy for eight emotion classes, including neutral. The techniques include seeding a Fisher Projection with the results of Sequential Floating Forward Search (SFFS-FP), which improves the performance of the Fisher Projection. The paper highlights the importance of physiological emotion recognition in various applications, such as medicine, entertainment, and human-computer interaction, and discusses the challenges and ethical considerations in data collection and analysis.This paper explores the development of machine emotional intelligence by focusing on the recognition of human affective states using physiological signals. The authors, Rosalind W. Picard, Elias Vyzas, and Jennifer Healey, argue that emotional intelligence is crucial for intelligent behavior and propose that machines should be capable of recognizing and responding to human emotions. They describe the challenges in obtaining reliable affective data and present a large dataset collected from a subject over multiple weeks, aiming to elicit and experience eight emotional states. The paper compares multiple algorithms for feature-based recognition of emotional states from this data, focusing on four physiological signals: electromyogram (EMG), photoplethysmography (PPG), skin conductance (SC), and respiration (R). The authors propose new features and algorithms to handle day-to-day variations in physiological signals, achieving an 81% recognition accuracy for eight emotion classes, including neutral. The techniques include seeding a Fisher Projection with the results of Sequential Floating Forward Search (SFFS-FP), which improves the performance of the Fisher Projection. The paper highlights the importance of physiological emotion recognition in various applications, such as medicine, entertainment, and human-computer interaction, and discusses the challenges and ethical considerations in data collection and analysis.