Please find the attachment for the donated "Speech" and "Instrumental
Sound" feature detection results on the search collection from DCU.
There are 176 files for each feature, respectively.

We provide the "Confidence" and "Relevance" value of a given feature
for each shot.  The confidence means how it is sure that the feature
is presented in the shot.  The relevance means how often or how much
the given feature is presented in the shot. "1.00" means the feature
is presented over the entire shot.

If a shot does not contain the given feature we simply set the
Confidence and Relevance to "0.00".

Our donation is of type B. We used Feature Development Collection for
tuning and testing our algorithm.  The algorithm was then run on the
Search Test Collection to get the final results.


Jiamin Ye
Postgraduate Lab
School of Computer Applications
Dublin City University
Dublin 9
Ireland