While doing some prior art search, I came up with the following two IBM patents: 7137062 (System and method for hierarchical segmentation with latent semantic indexing in scale space) and 7382933 (System and method for semantic video segmentation based on joint audiovisual and text analysis).
These patents are at the opposite ends of the “good patents rainbow”. Aside from the fact that audiovisual and text analysis was already studied and patents on similar subject matter were issued (see 6714909 from AT&T which deals with identifying segments based on several audiovisual parameters), the second patent also looks like a “give me something for my money” issue (there are “This invention was made with Government support […] awarded by the U.S. Army” and “The Government has certain rights in this invention” clauses).
The ‘062 patent is done by the book: a neat disclosure, references to major prior art documents, possible implementations, lot of mathematics behind every decision in the algorithm. Now this is what I call a good patent.
On the other hand, the ‘933 patent has a description of a computer for most of its disclosure and picture descriptions for the rest of it. This is what I call a not so good patent.
If only there was a way to stop the filing of the not so good patents and shrink the USPTO backlog…
Instant Blog 