Patient Room Monitor - Pre-evaluation

Introduction:

This project is intended for Hospitals in order to take care better of the patients.

Project description:

The system is for monitoring the health state of the persons and also their actions: gesture detection for opening the window, turning on/off the light, turning on/off the Air conditioning/Heating System, start/stop the television, switch the channels, increase decrease television volume, call the medical staff.

The system overview with the use cases is shown in the figure 1.

Hardware parts:

_______________________________

The required parts are:

x- Nexys Video FPGA platform

x- Pmod BT2 (Bluetooth transmitter receiver emulated on a UART interface)

x- Optical Sensors (two OV7670 image sensors)

x- Other sensors: e.g Heart beat rate, Respiration rate, ECG

x- Visualizing system: PC/Smartphone with Android

x- Wires and cables (Dupont, USB, HDMI)

The system block diagram with the information flow and main hardware components and tasks are shown in figure 2.

_______________________________

_______________________________

This is the pre-evaluation for the Digilent Design Contest 2017

Step 1: Software Blocks

Image acquisition:

For gesture recognition a stereo video stream is gathered from 2 OV7670 image sensors. Afterwards stereo matching algorithms are applied for stereo depth information.

2 video streams 640x480 VGA image

Health parameters: pulse, heart rate, respiration rate are gathered using ECG and heart rate sensors.

Step 2: Image Acquisition

Video streams for visual processing are fetched by two (2) Omnivision OV7670 cameras in RGB /YUV format.

Step 3: Processing Data

Image processing chain - the stereo image streams from the two sensor cameras OV7670 are analyzed using stereo depth algorithms, subsequent a feature extraction, segmentation and classification is done for prevaling the information about patient: hand, body, face.

The data from Heart and respiration sensors is processed using digital methods: filtering, information extraction algorithms: convolution, correlation.

Sobel edge detector filter was implemented, but more complicated filters are required for extracting the features: edge, contour and corner detection algorithms due to the fact that we want 3D depth information. As well Sobel offers modic results, more advanced filters like Prewitt or Canny offer much more information (Sobel is already implemented and it is recorded in the video attached to the next step), also more granular kernels would be beneficial due to the fact that OV7670 camera are low resolution VGA 640x480@30fps and the quality is 0.3 MP..

Step 4: Information Display

The useful information is displayed on a console display: 3D depth image , messages listed on the screen.

The Nexys Video board has video interfaces: HDMI or mini DisplayPort. Also the USB-UART/Ethernet interfaces and the Bluetooth/WiFi modules are used for sending information to a terminal application on the PC or on an Android Tablet/Smartphone.