A transformer-based network model for pitch detection

The web content introduces TuneNN, a network model that tracks the pitch of musical instruments. The author explains that the timbre of musical notes is influenced by various factors such as harmonic relationships, instrument resonant peaks, and structural resonant peaks over time. They highlight the use of different spectra, including the STFT spectrum, which accurately reflects harmonic relationships, and the Bark spectrum, which is a nonlinear compression of the STFT spectrum that aligns with human ear perception. The content also mentions the use of Cepstrum, CQHC, and MFCC features for pitch variations in speech. The author notes that TuneNN utilizes a transformer-based network model for abstract timbre modeling, supporting tuning for over 12 instrument types.

https://github.com/TuneNN/TuneNN

To top