Speech Phoneme-Enriched Audio Dataset (SPEAD)

The Speech Phoneme-Enriched Audio Dataset (SPEAD) represents a carefully assembled resource for researchers working in phonetic analysis, speech processing, and synthetic audio detection. This comprehensive dataset brings together high-quality speech recordings from 100 speakers representing diverse linguistic backgrounds, each contributing performances of specially designed phoneme-rich content that includes narrations, dialogues, and tongue twisters. The dataset was strategically developed to encompass the full spectrum of English phonemes and various articulation patterns, making it particularly valuable for researchers developing and testing algorithms in speaker recognition, speech enhancement, and audio deepfake detection systems. The inclusion of phoneme-dense material ensures that researchers have access to the rich acoustic variations necessary for building robust models.To provide additional analytical depth, the dataset also incorporates a contrasting element: 18 participants recorded passages with low phoneme density, creating opportunities for comparative studies of different speech patterns. This thoughtful design choice allows researchers to examine how phonetic complexity affects various speech processing applications. Privacy considerations were paramount in the dataset's development, with all recordings carefully anonymized while preserving the detailed metadata that researchers need for their work. SPEAD ultimately serves as an important benchmark resource, offering the phonetic variability and natural vocal expressions that are essential for advancing research in systems requiring a sophisticated understanding of human speech patterns.This dataset addresses a critical need in the research community by providing standardized, high-quality materials that enable meaningful comparisons across different studies and approaches in the field.