Skip to content

Recursive formula #5

@tucothien

Description

@tucothien

Hi, I'm trying to understand this algorithm.
I have a question about lines 76 and 77.
Line 76: logGamma_frame=np.log(a10/a01) + gMeanLogLRT + np.log(b01+b10/( a10+a00*np.exp(-logGamma_frame) ) )
As I understand, line 76 corresponds to the recursive formula [1](11), right?
But why logGamma includes the term np.log(a10/a01)? It looks similar to the final decision.
According to [1], the final decision is a10 / a01 * Gamma
But at line 77: probRatio[i]=1/(1+np.exp(-logGamma_frame))
Why do you use this formula and where is the final decision in [1](9)?

I have seen the same thing in vadsohn.m in matlab voicebox lib.
I'm struggling with it.
Please help me clarify it.
Thank you very much.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions