Speech emotion recognition (SER) has become an important area of research due to its wide applications in human-computer interaction, affective computing, and mental health. However, most existing studies focus heavily on adults, with limited attention given to children. This review highlights the urgent need to advance research on children’s SER, as detecting emotions in children poses unique challenges, such as limited labeled datasets, evolving vocal patterns, and the dynamic nature of childhood emotions. We examine both traditional feature-based approaches and recent developments in raw waveform modeling for SER. While handcrafted features have long dominated the field, very few studies have explored raw speech models despite their potential to capture nuanced emotional cues. Moreover, this review synthesizes SER research from 2014 to 2025, encompassing both adults and children to contextualize the current landscape. Notably, it devotes focused attention to children’s SER, outlining its distinctive challenges and key gaps that hinder progress. This child-centered perspective complements the broader discussion on adult SER, fostering an inclusive and age-diverse understanding of emotional speech modeling.