With the rapid advancement of artificial intelligence, intelligent media has seen significant progress, particularly in the area of music generation. This paper presents a comprehensive survey of the core technologies in symbolic music generation, reviewing specific applications of autoregressive models, Variational Autoencoders (VAEs), Generative Adversarial Networks (GANs), and diffusion models. We analyze over a dozen representative papers published in top-tier conferences (e.g., ICML, AAAI, ISMIR, IEEE) over the past two years, summarizing and comparing them across five key aspects: model architecture, data representation, input/output formats, control mechanisms, and practical applications. Our analysis focuses on how multi-modal inputs, emotional and stylistic tag control, and structured representations are used to enhance the quality and controllability (e.g., over dimensions like emotion and style) of generated music. Furthermore, we provide insights into future research directions in symbolic music generation. This paper aims to provide a systematic and valuable technical survey for researchers working at the intersection of artificial intelligence and intelligent media.