CQT-FW(Constant Q transform with Fixed length Window) is a time-frequency transform for multiple pitch recognition in audio signals(songs). The transform CQT(Constant Q Transform) provide analisys frequencies in a exponencial way, like music does, but it have problems with the time resolution. In CQT-FW we have the same frequency resolution as CQT, but with fixed length windows in time, what leads to a constant time resolution as well. And more, time and frequency resolution are not dependent like in CQT end DFT transforms. We also propose a post processing witch eliminate the high order harmonics, leaving only the fundamental frequencies. The transform got good results in a good range of frequencies, but still have problems in low frequencies.
Project Samples
