Databases (NTT, KEELE, and NoiseX) for Fundamental Frequency (Pitch) Extraction in Noisy Speech Signal.

The NTT database included speech samples from four male and four female speakers. Each speech signal consisted of approximately 149 10-second-long Japanese sentences sampled at 10 kHz with a 3.4 kHz band limit. From the KEELE database, we utilize five male and five female speech signals spoken in English. The total length of the ten speeches is approximately 6 minutes. The speech signals were sampled at a data rate of 16 kHz. Additionally, babble noise, train noise, and HF channel noise were extracted from the NOISEX-92 database.