This paper details progress during the first year of the JST/CREST ESP Project, on the creation of natural-speech databases for the analysis and synthesis of "Expressive Speech", and the development of software tools for parameter-extraction and speech database labeling. The research is still in its initial stages, but we now have a clearer understanding of the types of speech data that will be necessary, and of the software and speech processing tools that are available for analysis and treatment of the data. The testing of applications and prototyping in real-world situations is planned as future work, and our current task is the collection of a 1000-hour corpus of natural conversational speech upon which future research will be based.