Re: [PD] ANN - Gesture Recognition Capabilities
Hallo! in the description of the algorithm they talk about some pending patents... I guess they use more than simple ann comparison to get this working in realtime on a game console. Yes, I also would not use a feedforward neural network (as implemented in ann) for this. You have to use a model which also incorporates time. For example a hidden markov model or echo state network (= special kind of recurrent neural network) should work. (The second is currently the topic of my thesis - so maybe I will make some similar examples in summer, when I have some time again ... ;) LG Georg ___ PD-list@iem.at mailing list UNSUBSCRIBE and account-management - http://lists.puredata.info/listinfo/pd-list
Re: [PD] ANN - Gesture Recognition Capabilities
Hi, On Sun, 2008-05-04 at 08:37 +0200, Georg Holzmann wrote: in the description of the algorithm they talk about some pending patents... I guess they use more than simple ann comparison to get this working in realtime on a game console. Yes, I also would not use a feedforward neural network (as implemented in ann) for this. You have to use a model which also incorporates time. Hm. You can incorporate changes over time using a standard feedforward ANN by wrapping your time-ordered vectors over a given time period into a single input vector and increasing the number of inputs to the network accordingly. But of course this introduces latency and other problems (e.g. it could massively increase the number of training examples required). Pd has the ann_td external, which provides a 'time delay' neural network which I believe incorporates time using a method similar to that described above. For example a hidden markov model or echo state network (= special kind of recurrent neural network) should work. I'm intrigued! Presumably these approaches avoid the latency problem by maintaining the network's state? Are there other advantages -- easier to train? Jamie -- www.postlude.co.uk ___ PD-list@iem.at mailing list UNSUBSCRIBE and account-management - http://lists.puredata.info/listinfo/pd-list
Re: [PD] ANN - Gesture Recognition Capabilities
Hallo! Hm. You can incorporate changes over time using a standard feedforward ANN by wrapping your time-ordered vectors over a given time period into a single input vector and increasing the number of inputs to the network accordingly. But of course this introduces latency and other problems (e.g. it could massively increase the number of training examples required). Pd has the ann_td external, which provides a 'time delay' neural network which I believe incorporates time using a method similar to that described above. Yes of course, one other possibility is to use time delayed neural networks ... For example a hidden markov model or echo state network (= special kind of recurrent neural network) should work. I'm intrigued! Presumably these approaches avoid the latency problem by maintaining the network's state? Are there other advantages -- easier to train? Hm ... I did not think about latency ... but if you do not process the data in blocks there should not be a significant latency (also for the time delay NN) ? However, the advantage of the echo state network is that training is linear and you cannot get in a suboptimal solution as with feedforward neural networks (where the error surface has multiple local minimas) - see for example http://www.scholarpedia.org/article/Echo_state_network for a short introduction. And it is recurrent - so in general more powerful ... from the link above: On a number of benchmark tasks, ESNs have starkly outperformed all other methods of nonlinear dynamical modelling If you are interested, I implemented ESNs (with various extensions) in a C++ library with python bindings: http://aureservoir.sourceforge.net/, a PD external will hopefully follow in summer ... LG Georg ___ PD-list@iem.at mailing list UNSUBSCRIBE and account-management - http://lists.puredata.info/listinfo/pd-list
Re: [PD] ANN - Gesture Recognition Capabilities
Hi all very interesting, really.. yet it does take me away more and more from my inital question ;-) And a practical solution to implement some simplicistic gesture recognition in PD.. ..and wait for the summer, when the PD bindings are ready... I hope there will be some creative, artisitic examples then.. Best Regards Luigi Am 04.05.2008 um 13:59 schrieb Georg Holzmann: Hallo! Hm. You can incorporate changes over time using a standard feedforward ANN by wrapping your time-ordered vectors over a given time period into a single input vector and increasing the number of inputs to the network accordingly. But of course this introduces latency and other problems (e.g. it could massively increase the number of training examples required). Pd has the ann_td external, which provides a 'time delay' neural network which I believe incorporates time using a method similar to that described above. Yes of course, one other possibility is to use time delayed neural networks ... For example a hidden markov model or echo state network (= special kind of recurrent neural network) should work. I'm intrigued! Presumably these approaches avoid the latency problem by maintaining the network's state? Are there other advantages -- easier to train? Hm ... I did not think about latency ... but if you do not process the data in blocks there should not be a significant latency (also for the time delay NN) ? However, the advantage of the echo state network is that training is linear and you cannot get in a suboptimal solution as with feedforward neural networks (where the error surface has multiple local minimas) - see for example http://www.scholarpedia.org/article/Echo_state_network for a short introduction. And it is recurrent - so in general more powerful ... from the link above: On a number of benchmark tasks, ESNs have starkly outperformed all other methods of nonlinear dynamical modelling If you are interested, I implemented ESNs (with various extensions) in a C++ library with python bindings: http:// aureservoir.sourceforge.net/, a PD external will hopefully follow in summer ... LG Georg ___ PD-list@iem.at mailing list UNSUBSCRIBE and account-management - http://lists.puredata.info/ listinfo/pd-list --- Luigi Rensinghoff [EMAIL PROTECTED] skype:gigischinke ichat:gigicarlo ___ PD-list@iem.at mailing list UNSUBSCRIBE and account-management - http://lists.puredata.info/listinfo/pd-list
Re: [PD] ANN - Gesture Recognition Capabilities
nice! in the description of the algorithm they talk about some pending patents... I guess they use more than simple ann comparison to get this working in realtime on a game console. for basic gestures it might work, but you have a 3d space (which means a lot of data) and no clear start and end trigger for your movements, so to me this looks like a challenging task... but if you manage to somehow calibrate your data before you send it to ann (maybe normalize x, y, z and time? or use statistic methods to get more meaningful values, also think how a gesture's start and end can be found) then ann hopefully does the rest. marius. Luigi Rensinghoff wrote: Hi List... http://video.google.com/videoplay?docid=-2991739314359132538 Maybe a question to who ever used ann, or maybe the two developers Johannes Zmoelnig and Davide Morelli I posted a link... Do you think ann is powerful and fast enough to do something like that ?? Would be nice to hear your opinion about that Thanks Luigi ___ PD-list@iem.at mailing list UNSUBSCRIBE and account-management - http://lists.puredata.info/listinfo/pd-list ___ PD-list@iem.at mailing list UNSUBSCRIBE and account-management - http://lists.puredata.info/listinfo/pd-list
