The plan, at this point, is to use a gammatone dictionary and sparse coding to train more efficiently on the TIMIT dataset. Gammatone functions are filters applied on frequencies. The goal is to keep frequencies processed by our ears. Joao explains this greatly in his blog. The idea is to use this sparse-coded version as input of a spike and slab RBM segmented by frames of 160 samples and with the previous, current and next phones. I kept the parameters use.

Experiments in speech synthesis with neural networks by Pierre-Luc Vaudry. Hubert Banville successfully implemented the kind of architecture I also had in mind. In this experiment, I wanted to see if having this standard deviation predicted by the neural network could improve the results.

Suatu hal yang tak bisa diulang, tak bisa dibagi, tak bisa dipinjam, tak bisa dibeli, tak bisa ditambah, tak bisa dipercepat, kadang kalau direnungi bisa saja menetes air mata. Sampai saat ini waktuku terbuang sia sia. Maaf Tuhan aku tak tahu apa saja yang telah kuperbuat sampai aku sangat menyesal saat ini. Tolong, aku ingin hidupku lebih efektif, produktif, bermanfaat bagi keluarga, bangsa ,dan almamater. Persaingan antar penjual memang makin memanas di dunia yang seperti ini, serba ada-serba modern.

This blog has no articles. Subscribe to my blog! Post to my blog.