Bestand wählen
Merken

NuFaucet [DEMO #6]

Zitierlink des Filmsegments
Embed Code

Automatisierte Medienanalyse

Beta
Erkannte Entitäten
Sprachtranskript
no no pontiff who wanted to know if you didn't have the structures that about
half of the area but will bring result on whether you then the comparison word we use the most per-capita compared to everybody else in the world to me I found that people use and 1 of the things that is all that is the chief except Canada which if it were pretty helper Petrovic needed if we ever want wonder
we conserve water we really can't have we have no idea how much water were using predictors all any because no ever wanted to know all our article basement take my meter reading every day or have away from my own management utility coming to give me a reading every 3 months and upon York City have found out that he has paid by the frontage so harbor why the building is and you have a limited amount of water and that's how it gets very we so so so you are measuring how much water using put it would mean a likelihood from in there would you some would miss some of them utility the company will have each 1 of these neuroelectrical ones digital once but day requires them doing the readings and the data is available to you you can do your own barbarianism about but if you do that you have to have bad and it's expensive so that so and I'm sure you
guys are aware of this so this picture in this region of times so what is your view of it can measure those 4 Fossetts without putting in a photometer because if you are there is no what it sounds like so each 1 of
us has has its own unique so so the question is can can teach you pay to want to detect this out until me that is that which fosters being used eventually wanna know what the duration is in which users because you have their own unique characters if you so that's what would be the running of in that room in the back of
the interns recordings so you see this 1 this is the let's say the hot water you see if this is the flow it's pretty loud and you this is the number and this is actually is little different is like this that f of the of the 50 so this not and that's actually something I have to implement eventually to to reduce the amount of data so this is there are the less income water the clicking sound it's them directly lincat these was will lose but there was a list you want to use you identify which velocity is the actual sentences pay it's just recorded with a little lower like and it's going down to the right sink hot water lines so Reggie listening through the feline Davila here which lasted that water and could copper actually come to and this 1 is the right thing because of the settlement renowned aligned with Mike is easy it's kind allowed so the trained and then this project actually space off matzo Matt's projects are them then you pick pretty so to me want understand how to get the word so basically what I was of the initial training work very well defined by the need to do a lot of data disputed through so that end up running at 10 Paul sets and then had to that 10 times and what I was doing these data generation and thing if I disturb somebody yesterday was sitting there just keep hitting the the fastest because the person was waiting inside the automatic the map so out of the 1st set of data from the lab sink hot water at a red the called the 2 codes the right color because that's a political 1 and 1 whole bouncing back here to this is the difference of these anomaly output from the new big and their assault that any given the detection saying that's not normally be seeing a non anomaly here because new Big Data the gap because water the water would assist the training I have pulses of have a lot of flat lines between the effect of lies models in section 2 so Ford actually covered work would need to detect the data and then take that out of the equation then let's start this underlying needed to see if I can see that the tag you know something I have this so frozen data into it that decided to do the right the right half which is right here it's actual out and we ran a radical in the hearts bed it decided that it didn't want it didn't know that that 1 wasn't something in New before so we give a red the knowledge and then the left cold actually became clear on both sides but what I found this depends on the actual data itself for example these guys for you train want dataset training on the left coast the amplitudes really small so this is not enough data for that system to learn and then if it did short like here on this 1 this 1 this was trained on these guys 10 of those individual 100 but still not enough data from you pick the learn so that's something to resolve and that's pretty
much as water in the do next better feature extraction results which just pumping way too much data through sitting at each summer doing learning so that over an hour sitting there waiting for the computer turned through so I I I think using a structural model pull more data more feature out as a reduced amount of time in each process learn and appeal and also for better detection lead to operate in multiple domains edges just temporal loss of frequency I need to run in multiple both combination of them just to get the corrugated station and that's at any questions yes for the training error using the Python implementation the C implementation but on it's strictly has called people visually look into this sea because this lot of flat file that has 6 that there is some like it's so slow can all this morning number so the inside of the we would amend words only here already yeah that's right right and of course this book full of people in the area which sounds section could be vague enough to detect other things than just fastest because I mean like alarm system threat if you think about it they're really just anomalies that happened in the Dutch we want this this is the bigger project is a project of this is losing their intersection and I think there is a rather than line around the and a time and amplitude i should be 0 tell you know its neighbors and then all notified as of today is generally you don't wanna bang eigenvalue neighbors doors and you too noisy you or somebody else to review and and maybe this so he left field but could be used the sound attention can be used to detect the sound the fire started to be used as a sort like ultrasensitive sounds to use it will sound we will be the virus itself doesn't have to much disturbance and there's something to acquire parameters for the noisy neighbor because we're using them in a force membrane and then the pipes the conduit Holger detector on the fires and so it's an interesting question as well that the reason I ask is because a lot of our low-income housing don't have you know fire detection or they usually very low enough implementation loser maybe can be used to detect the changes of the sounds in the overall structure in particular detected it's just so hard and so determines the means of personal through and that's how we get there is not a good time to know that look at the blood normal early before yes or set of visually impaired people I mean walking finding I think if you can get predictability acoustic patterns you can really help safety wise because such as something use thinking about all the acoustic because it has to do with this feature these guys
ability another here the higher orders of the does go up and down it's kind of like that but this feature ability of all the feature the user some of new ones and sort of acoustic speak involves and in exactly use of 4 additional features directions and always thinking like sounds that overlap during the likes of things that stick out those are what this has really good application for because those kind like things that are unusual a car coming at you or like street like an intersection like dangerous because it really is the involvement of users have yourself will listen to allow thing and because you by that that that would require system implementation of foreign that's it I would just below was affected you found out that
I had my my my my
Resultante
Einfügungsdämpfung
Gleichungssystem
Computer
Gerichteter Graph
Raum-Zeit
Computeranimation
Richtung
Eins
Prognoseverfahren
Puls <Technik>
Datenmanagement
Mustersprache
Meter
Vorlesung/Konferenz
Gerade
Funktion <Mathematik>
Addition
Parametersystem
Sichtenkonzept
Krümmung
Gebäude <Mathematik>
Temporale Logik
Ausnahmebehandlung
Frequenz
Quelle <Physik>
Generator <Informatik>
Datenfeld
Forcing
Menge
Rechter Winkel
Garbentheorie
Projektive Ebene
Ordnung <Mathematik>
Lesen <Datenverarbeitung>
Fehlermeldung
Geschwindigkeit
Eigenwertproblem
Computervirus
Subtraktion
Wellenpaket
Klassendiagramm
Wasserdampftafel
Mathematisierung
Schaltnetz
Zahlenbereich
Implementierung
Domain-Name
Datensatz
Informationsmodellierung
Arbeitsplatzcomputer
Inverser Limes
Datenstruktur
Hilfesystem
Demo <Programm>
Soundverarbeitung
Folientastatur
Softwarewerkzeug
Mailing-Liste
Paarvergleich
Physikalisches System
Elektronische Publikation
Datenfluss
Quick-Sort
Mapping <Computergraphik>
Flächeninhalt
Codierung
Wort <Informatik>
Kantenfärbung
Stab
Lie-Gruppe

Metadaten

Formale Metadaten

Titel NuFaucet [DEMO #6]
Serientitel 2015 Spring NuPIC Hackathon
Anzahl der Teile 19
Autor Yeh, John
Lizenz CC-Namensnennung 3.0 Unported:
Sie dürfen das Werk bzw. den Inhalt zu jedem legalen Zweck nutzen, verändern und in unveränderter oder veränderter Form vervielfältigen, verbreiten und öffentlich zugänglich machen, sofern Sie den Namen des Autors/Rechteinhabers in der von ihm festgelegten Weise nennen.
DOI 10.5446/18057
Herausgeber Numenta Platform for Intelligent Computing (NuPIC)
Erscheinungsjahr 2015
Sprache Englisch

Inhaltliche Metadaten

Fachgebiet Informatik
Abstract Doing audio WAV analysis on faucet sounds to identify which faucet was turned on.

Zugehöriges Material

Ähnliche Filme

Loading...
Feedback