search

shunsukeaihara / pyssp

python speech signal processing library
MIT License
18 stars 7 forks source link

Threshold computation #1

Closed maelp closed 7 years ago

maelp commented 7 years ago

The computation should be (lambda0 - lambda1) / (e0 - e1) + lambda0 - (lambda0 - lambda1) / (1 - e1/e0) in https://github.com/shunsukeaihara/pyssp/blob/master/pyssp/vad/ltsd.py#L50, right now you use e0 / e1

shunsukeaihara commented 7 years ago

Thank you for point out. I fixed the threshold calculation by referring to eq.3 in the following paper.

Ramırez, Javier, José C. Segura, Carmen Benıtez, Angel De La Torre, and Antonio Rubio. "Efficient voice activity detection algorithms using long-term speech information." Speech communication 42, no. 3 (2004): 271-287.

http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.102.9761&rep=rep1&type=pdf