For many days, I was wandering the physical meaning of BER (bit-error rate) and SNR (signal to noise ratio), which are heavily used by designers in communication domain. Besides, the effectness of ECC (error correction code) was far from illustrative.
Therefore I built this task (in Matlab and C) to monitor the physical meaning of SNR, BER, ECC (LDPC), illustrated in the context of (Image) video transmission.
The Matlab GUI simulates video transmission via a gaussian noise (AWG) channel. The modulation sheme is QPSK. For error correction, a LDPC decoder is utilized.
To speedup the simulation, the model uses a custumized LDPC encoder/decoder, rather than the built in one form communication_toolbox. This brings ~100 speedup. The LDPC encoder/decoder is compatible to IEEE 802.11ad starndrd, by default it is in 1/2 coding rate, you can hack it to use 5/8, 3/4, or 13/16 coding rate. The block size of the code is fixed to be 672 bits. It takes logrithmetic input from the mapper. The LDPC decoder is implemented in fixed-point to represent industrial chip behavior. The bitwidth is between 4-10. Currently we only provide mex code to be used in Matlab. The original c code, as well as the exact bitwidth, are still confidential. We hope to open-source to the public soon.
- 64-bit operating system of ethier Windows, Unix, or MAC OSX
- 64-bit Matlab, with licences of:
- basic Matlab license
Download this git page
Otherwise, you can simply fork this page by:
git clone https://github.com/yanxianghuang/video_ldpc
Run the script from Matlab:
Open Matlab, entering the src folder, start the program by typing
in your Matlab commandline. A figure template should appear.
Adjust the parameters in the right-top corner Panel:
- select file for testcase, you can choose the one in '../testcase/sample.avi';
- choose the SNR value in dB for simulation, preferably chose 3;
- for 'start from' item item, you may choose 0, meaning we start from the first frame (the 0 second).
- tick the Updating new frames, if you want the simulation process to update new frames automatically (streaming).
- click Simulate, you are ready to see the effect, of the original frame(s), corrupted by the signal frame(s), and the decoded form error correction code frame(s).
- To stop, untick the 'updating new frame', it will stop when finishs current frame. To start a new simulation, click Simulate button.
Be patient as the simulation may take long. This is mainly due to the time-consuming demapper function from communication_toolbox. I hope one day I'll find the time to convert it to Python, or reuse my own demapping, for speed-up. Enjoy, and if you have and suggestions, please shot.
You can modify the program as you like. For instance, you can change the modulation scheme, channel noise model, LDPC schemes (or even substitute it with one you have).
- show_pic_ber.m (the main function)
- show_pic_ber.fig (the GUI interface, it will be open once show_pic_ber.m is used)
- load_video.m (loading video function/script)
- pic2bit.m (function: convert picture/frame into an array of bits)
- bit2pic.m (function: reverse of pic2bit)
- encoder.(mexw64/mexa64/mexmaci64) (LDPC encoder binary files in Windows, Unix, MAC OSX)
- decoder.(mexw64/mexa64/mexmaci64) (LDPC decoder binary files)
- test.avi (testing video)
- ldpcsnr3.jpg (example screenshot)
The main file contains the gui_load_video function, which is the main processing block, It is executed everytime the start-Butten was hit.
It first get the data from a frame, pack it into a encoder blocks. On each blok, encode, modulation, mapping (demodulation), and decoding. Once all blocks of a frame is done, paint the result on GUI. And if updating_new_frames option is selected, it will continue to the next frame. Otherwise, the program will stop.