| Popis: |
This investigation explored the relationship between speech tokens’ temporal and spectral characteristics and blind reverberation time estimation accuracy. The role of two factors determining these characteristics: phonetic contents, and rate of speech and effort used during token articulation, was assessed. To this end, a corpus of logatome recordings was artificially reverberated using 160 simulated and 16 measured room impulse responses, whereupon reverberation time estimates were acquired using a state-of-the-art algorithm based on the statistical model of late reverberation. Results of analyses indicate that both factors affect the accuracy, and should be considered when developing more robust reverberation time estimation algorithms. |
