search

TWOEARS / blackboard-system

Two!Ears Auditory Model - Blackboard system module
http://docs.twoears.eu/en/latest/blackboard/
GNU General Public License v2.0
3 stars 2 forks source link

A KS for identifying a target after segmentation #17

Closed kashefy closed 8 years ago

kashefy commented 8 years ago

This adds a KS that mimics the IdentityKS except that it applies the mask generated by SegmentationKS to the AFE Signals before inference.

A field is added to the SegmentationHypothesis for storing the hop Size of the Segmentation mask for each source. This is needed to determine the temporal resolution of the mask in case it needs to be re-sampled to match the temporal resolution of the AFE Signal.

Spectral features are excluded from masking.

Dependent on TWOEARS/auditory-front-end#4

hagenw commented 8 years ago

Cool, could you please also make a branch identify_segment in TWOEARS/documentation and add a short section describing the new knowledge source under the general section Identification knowledge sources

kashefy commented 8 years ago

@hagenw documentation done (TWOEARS/documentation#8), thanks for pointing it out.

hagenw commented 8 years ago

The documentation is now included, see http://twoears.aipa.tu-berlin.de/doc/latest/blackboard/knowledge-sources/#segment-identity-knowledge-source-segmentidentityks

In order to see how this new knowledge source is working it would also be cool if you could add a new folder containing an example scene to TWOEARS/examples, so everyone can test it.

Otherwise @chrschy should look at your comment above and after that it should be ready for merging.