In this presentation on Compressed Sensing enabled video streaming for wireless multimedia sensor networks we will look at a networked system for joint compression, rate control and error correction of video over resource constrained embedded devices. This theory is based on the theory of Compressed Sensing (CS). Solution to the problems such as encoder complexity and low resiliency to channel errors need to be addressed. Compressive Distortion Minimizing Rate Control, regulates the CS sampling rate, the data rate and the rate of a simple parity based channel encoder are the methods used to address this issue.
- Networks of wirelessly interconnected devices
- Retrieve multimedia content such as video and audio streams, still images from the environment.
- Can store, process in real-time, correlate and fuse multimedia data originated from heterogeneous sources.
- Enhance existing sensor network applications such as tracking, home automation, and environmental monitoring.
This Signal processing technique is used for efficiently acquiring and reconstructing a signal. The entire signal can be determined from relatively few measurements. Not acquiring that part of the data that would eventually just be ‘thrown away’ by lossy compression. There is a faithful recovery of signals. Using M<<N measurements, N is the number of samples required for Nyquist sampling. It can offer an alternative to traditional video encoders. It enable systems that sense and compress data simultaneously. There is a low computational complexity for encoder
Architecture Of Compressive Distortion Minimizing Rate Control
Components of the system are CS Camera, CSV Video Encoder, Rate Controller and Adaptive Parity Block. It's working is as given below
- Takes a sequence of images at a user-defined number of frames per second
- Wirelessly transmits video encoded using compressed sensing
- Congestion control and protection against channel losses are provided
- Combines functionalities of application layer, transport layer and physical layer
- Compressed sensing image capture takes place.
- Can be either a traditional CCD or CMOS imaging system, or it can be a single-pixel camera.
- The samples of the image are directly obtained by taking a linear combination of a random set of the pixels.
- Samples generated are passed to the video encoder.
CS VIDEO Encoder
Receives the raw samples from the camera and generates compressed video frames.
- Input end-to-end RTT of previous packets and sample loss rate
- Determine optimal sampling rate for video encoder
- Rate control law maximizes video quality
Adaptive parity block
- Determines a parity scheme for encoding the samples
- Uses measured or estimated sample error rate of the channel
- Input from the video encoder
Represent each frame of the video by 8-bit intensity values. Image can be sampled using a scrambled block Hadamard ensemble, y = H32 * x where y represents image samples, H32 is the 32 X 32 Hadamard matrix and x the matrix of the image pixels.
RATE CONTROL SUBSYSTEM
- Provides fairness in terms of video quality
- Maximizes the overall video quality
- To avoid network congestion, sender needs to regulate its rate to allow any competing transmission at least as much bandwidth it needs. Sender needs to regulate its rate to make packet losses due to buffer overflows reduced.