Project
# | Title | Team Members | TA | Documents | Sponsor |
---|---|---|---|---|---|
83 | Room Occupancy Sensing |
Aakarsh Sethi Steve Wang Yohann Puri |
Eric Clark | design_document0.pdf final_paper0.pdf proposal0.pdf |
|
Team members Aakarsh Sethi - assethi2 Steve Wang - spwang3 Yohann Puri - ypuri2 The basics The basic idea is to create a prototype for counting the number of people entering and leaving a room. This is achieved by taking snapshots of contact impressions left on a mat along with using a coupled set of IR sensors which are triggered when a person crosses a line allowing counting. The impression snapshots would help distinguish between two classes, humans and objects. This combined with counting done by IR sensors(Increment every time two sensors are crossed by an object) helps us get a pretty accurate count of the number of people who have entered and left a room. An FPGA would be used to aggregate sensor data and a machine learning model would be created to analyze this data. The results we are aiming for would not be time bound. Paper that gave us inspiration for this fabric http://onlinepresent.org/proceedings/vol87_2015/1.pdf http://ame2.asu.edu/projects/floor/papers/srinivasanp_pressurefloor.pdf http://www.nime2011.org/proceedings/papers/L01-Roh.pdf Instructable explaining how its made - http://www.instructables.com/id/Flexible-Fabric-Pressure-Sensor/?ALLSTEPS Materials for pressure sensing fabric - https://www.lessemf.com/fabric1.html - Conductive material https://www.adafruit.com/products/1361 - Velostat/Linqstat Modules: Floor Sensing 36in x 24 in carpet for initial prototype with 2 sensors per square inch, totalling to 360 sensors. We will be using custom made fabric force sensors spaced evenly beneath the surface of the carpet. The sensors will be made by sandwiching a pressure-sensitive conductive plastic between two layers of conductive fabric. The resistivity between the two conductive fabrics will change as force is applied to it. The sensors will be connected to an FPGA that will serialize the data and pass it to an onboard computer. IR Sensing On the edges of the carpet, we will be using IR proximity sensors to determine the number of people that have entered the sensing area. The IR sensors are coupled such that when a person crosses the south lining of the mat, one IR sensor gets triggered but the person is only counted if he/she crosses the north lining of the mat. This is detected by the second IR sensor on the other end. Machine Learning The idea is to use the snapshots of the mat that the FPGA delivers, to use machine learning to classify whether the object on the map is a foot, wheel, box, etc. The mat would have contact sensors which would deliver a silhouette of what is on the mat. Shoes will have a certain range of shapes as opposed to mail carts and moving trash bins. raining data would include different types of shoes. Training would be supervised. The proximity sensors help give a count of how many objects went across the mat and the analysis of the snapshots tells us whether what went in or out was a person or an object. |