Speech recording protocol and quality standardization project: An overview

This presentation provides an overview of the sp-recPAQS (Speech Recording Protocol and Quality Standardization) project, which was launched to address these practical matters. The project aims to classify speech recording quality into four classes, define recording protocols and quality criteria for each class, and develop support tools to achieve the best possible recording quality within each class. Its ultimate goal is to establish guidelines that promote the accumulation and reuse of speech data for both research and practical applications. Recording high-quality speech data requires selecting high-performance equipment and following strict procedures, which makes the process demanding. Even when aiming for recordings that meet only minimum requirements for a specific research purpose, researchers often struggle to determine appropriate equipment and procedures. In addition, speech recording requires considerable resources, including speaker arrangement, recording time, and preparation. Therefore, it is beneficial if the recorded data can be reused by others. This presentation introduces practical procedures for recording reusable speech data in typical environments where background noise and reverberation cannot be ignored. It also discusses unavoidable signal changes caused by environmental and instrumental noise, reverberation, and A/D conversion, as well as the importance of microphone sound pressure calibration and approaches to minimize these effects under real-world constraints.