Case study: Apple — Recognising hand gestures

AmberScope is a patent searching engine developed by Ambercite which is able to find patents that may be missed by existing patent search processes.

AmberScope is introduced in this video, but how well does AmberScope work in practice?

In this case study, we show how the use of filters can help make a patent search in a complex area both manageable and effective.

In Case study: Basic principles — Google glasses, we look at the ability to find potential prior art for a Google patent covering glasses for augmented vision. Now it is the turn of an Apple patent to be placed under the prior art ‘lens’ of AmberScope.

In 2011, Apple was granted a US patent US7877707 for a method of detecting hand shapes on a multi-touch screen, with a priority date of January 2007.

The granted patent includes a list of prior art document, but did the examiner for this patent identify all of the relevant prior art?

This is where AmberScope can be very useful. The image below shows what happens if we run a search for this patent US7877707 (‘707) in AmberScope.

Each dot represents a patent, with the dot surrounded by the red circle being ‘707, which was the ‘focus ’ patent of this patent network. All of the patents connected to the ‘707 patent are shown, and these are represented in a light grey colour. Some other features are:

  • Purple lines show backward citations, and green lines forward citations.
  • The question mark in the patent box shows that the patent has not yet been rated by the user. AmberScope includes the facilities to capture personal ratings on any patent, and this can be very handy for future referral. Currently, patents can be rated from 0 to 4.
  • The ‘707 patent has an AmberScore rating of 1.8. AmberScore is a proprietary algorithm that considers the influence of the patent in the network. An AmberScore value of 1.8 is lower than average – the average granted US patent over the last 20 years has an AmberScore value of 1. The dot size on the screen is proportion to its AmberScore rating.
  • The image also shows some of the highly rated patents that are connected to the patents that are connected to the ‘707 patent. These indirectly connected patents could be regarded as influential ‘friends of friends'. We refer to these patents as ‘ghost ’ patents, and they are mostly shown as greyed out in this case.

Ghost patents can easily be identified as patents with connection lines overlying them ,

as opposed to the connection lines hidden behind the dots 

Ghost patents can be very valuable, as these can disclose inventions that were not considered by the patent examiner (otherwise they would be listed in the search report) but still may be relevant.

But there are a lot of patents on the screen, almost too many to look at. So we made some simplifying if probably risky assumptions, in this case:

  • Relevant prior art was not filed by Apple. We made this assumption because Apple should have declared all of the prior art it knew about to the US patent office, and this should have included earlier Apple patents
  • The prior art would have been filed quite recently, say in the year 2000 or afterwards. It after all covers touch screens, a relative recent development. And being prior art, it would have been filed in 2007 or before.

These assumptions are risky, but can be justified in this case because this is only a worked example of a patent search.

Practically, these assumptions are entered into the filter box.

Which leads to the following image, containing fewer patents to review (the Apple patents have a green circle around them, which makes them easy to avoid).

In addition we can make a third assumption, namely that relevant prior art might be that which the examiner did not know about, which are the ghost patents. We can make this assumption because the examiner made the decision to grant the patent based on the prior art they knew about, and so it might harder to argue that the patent is invalid based on this already known prior art. And a fourth assumption, which is that if we are looking at the ghost patents, we should start with the biggest dots first, as AmberScore has told us that these likely are the most important and relevant patents.

Which suggests that we should start with the big ghost node on the far left of the screen, which happens to be US6888536 and has an AmberScore value of 11.

And In fact, when we look at it, it is quite relevant, disclosing:

Apparatus and methods are disclosed for simultaneously tracking multiple finger and palm contacts as hands approach, touch, and slide across a proximity-sensing compliant, and flexible multi-touch surface

So already, by just making a few assumptions, we have found what could be a relevant patent. But what happens if we make this the focus patent, as the patents connected to this might be very relevant?

It is quite a crowded landscape as you can see, and which we would expect from a high AmberScore value. But let’s apply the same assumptions as before, namely narrowing our search to patents filed between 2000 and 2007 and highlighting the Apple patents, and see what happens:

Better, but still a little bit crowded. So we might make another assumption, that a relevant patent may be among the highest rated of the patents on the screen. But there is no need to specifically look for ghost patents, as we can assume that many of these patents will not be directly connected to the original Apple ‘707 patent.

How about we hide say the bottom 60% of the patents using the ‘Percentage ’ filter? (i.e. hide the bottom 60% of patents as ranked by AmberScore) By doing this, we can remove most of the patents, as shown in the image below.

And maybe we can start exploring these patents — as there is a manageable number. As we do so, we discover that one of these patents is very interesting, being US7618323 which covers:

A gaming machine has a processor for conducting a wagering game on the gaming machine and a gesture-sensing mechanism. The gesture-sensing mechanism can be used for providing various inputs


And we could keep going from here, perhaps change our assumptions, and explore every patent in these networks. This is exactly what you should do if you are doing a serious prior art search.

So does US 6888536 or US7618323 invalidate the apple patent ‘707? I will leave this for other people to judge, but I can see how they might be relevant.

But why did the patent examiner for the ‘707 patent miss either of these patents? This becomes clearer if we explore some of the key concepts usually used to search patents on, which are listed in the table below.

So while there are keywords in common, including in the title, the IPC and US patent classifications differ completely for all three patents. But if this is the case, why did AmberScope find these patents?

There are two main reason why AmberScope is so effective at finding patents

  1. The first reason why is that AmberScope draws upon the collective intelligence of all of the examiners (and patent applicants) that examined the patents in these networks. Each examiner would have used a slightly different strategy to search for patents, and so would have found slightly different results. By combining these different results, AmberScope is able to find and highlight patents missed by individual examiners.
  2. The second reason is that AmberScope applies principles of game design to help motivate the searcher and lead them to their final outcome, increasing their productivity and outcomes. Users tell us that AmberScope is fun to use compared to other patent searching engines - and this helps them achieve their desired objectives.

And this is why AmberScope is so effective at finding patents missed by conventional patent searching techniques.

 I spent two days looking for a relevant patent for a client using a conventional search. The second patent took me 30 minutes to find with AmberScope.

R. W., Patent Attorney at Griffith Hack

 My role is directly analyzing patents including claims, infringement and prior art. AmberScope appears exceptional for identifying prior art from anything else I’ve seen.

Paul Morinville, CEO at OrgStructure US