Phase modelling with Neural Nets
 
 
Go to file
David e01a050516 broken... 2019-11-22 15:28:14 +10:30
README.md updated README 2019-11-22 12:42:55 +10:30
phasenn_test1.py moved out of codec2 repo 2019-11-17 12:05:58 +10:30
phasenn_test2.py moved out of codec2 repo 2019-11-17 12:05:58 +10:30
phasenn_test3.py moved out of codec2 repo 2019-11-17 12:05:58 +10:30
phasenn_test4.py moved out of codec2 repo 2019-11-17 12:05:58 +10:30
phasenn_test5.py moved out of codec2 repo 2019-11-17 12:05:58 +10:30
phasenn_test6.py moved out of codec2 repo 2019-11-17 12:05:58 +10:30
phasenn_test7.py moved out of codec2 repo 2019-11-17 12:05:58 +10:30
phasenn_test7a.py moved out of codec2 repo 2019-11-17 12:05:58 +10:30
phasenn_test8.py moved out of codec2 repo 2019-11-17 12:05:58 +10:30
phasenn_test9.py test9 n0 scaled by P_max, sampling excitation on width length DFT bins rather than m*Wo 2019-11-22 12:41:36 +10:30
phasenn_test9a.py broken... 2019-11-22 15:28:14 +10:30
phasenn_test9b.py using e_rect as input (ex phase) rather than phase_rect (ex+dispersive) 2019-11-22 12:59:33 +10:30
phasenn_test10.py work in progress combination of n0 est and dispersive part 2019-11-22 12:43:59 +10:30
phasenn_train.py moved out of codec2 repo 2019-11-17 12:05:58 +10:30

README.md

PhaseNN

A project to model sinusoidal codec phase spectra with neural nets.

Recent breakthroughs in NN speech synthesis (WaveNet, WaveRNN, LPCNet and friends) have resulted in exciting improvements in model based synthesised speech quality. These algorithms typically use NNs to estimate the PDF of the next speech sample using a history of previous speech samples. This PDF is then sampled. As such, speech is generated on a sample by sample basis. Computational complexity is high, although steadily being reduced.

Speech codecs employing frequency domain, block based techniques such as sinusoidal transform coding can deliver high quality speech using block based synthesis. They typically synthesise speech in blocks of 10-20ms at a time (e.g. 160-320 samples at Fs=16kHz) using efficient overlap-add IDFT techniques. Sinusoidal codecs use a similar parameter set to NN based synthesis systems (amplitude spectra and pitch information).

However for high quality speech, sinusoidal codecs require a suitable set of the sinusoidal harmonic phases for each frame that is synthesised. This work aims to generate the sinusoid phases from amplitude information using NNs, in order to develop a block based NN synthesis engine based on sinusoidal coding.

Status (Nov 2019)

Building up techniques for modelling phase using NNs and toy speech models (2nd order filters) in a series of tests.