<p>Speech data is foundational to advancing both speech research and technology applications. Data collection and annotation are critical processes that ensure the usability and effectiveness of speech data. Despite growing academic and industrial speech databases, persistent challenges in data collection and annotation exist, including limited experience sharing and a lack of standardized scheme design, especially for underrepresented populations or low-resource languages. Therefore, this paper first outlines the overall framework of data collection and annotation and the key components with representative methods and commonly used tools. Furthermore, this paper showcases the tools and platforms we build for collecting and annotating speech of second language acquisition, hearing-impaired children, and mental health monitoring. Finally, we discuss the emerging trends, challenges, and possible solutions in the era of big data and large models. By integrating methodological guidelines with empirical case studies, this paper provides actionable insights for constructing speech databases.</p>