Sequence to One RNN for voiceprint recognition

Element-Research / rnn

Recurrent Neural Network library for Torch7's nn

BSD 3-Clause "New" or "Revised" License

939 stars 313 forks source link

Sequence to One RNN for voiceprint recognition #251

Open ForeverGaara opened 8 years ago

ForeverGaara commented 8 years ago

hello @nicholas-leonard

I am very interested in Sequence to One RNN, woud it be possible to apply RNN on voiceprint recognition/speaker recognition, and what kind structure of RNN can be used to do this task. looking forward your reply, thank you.

nicholas-leonard commented 8 years ago

Yes. You can checkout our sequence-to-one.lua script. The key is to use a SelectTable(-1) after your Sequencer(RNN).

Trojahn commented 8 years ago

Hi @nicholas-leonard!

How can i change the example script to: 1 - Support 1D vector values instead of simple numbers? (Each sample have the same size) 2 - Support sequences of variable size? (Even considering batch_size = 1)

My aim is to classify if a given sequence (formed by a bunch of 1D vectors values) is either of a given class or not (binary classifier).

Thank you!