*Corresponding author: Jun Wu
Hardware Implementation of a Pseudo-Analog Wireless Video Transmission System
Shi Chen, Jun Wu*, Haoqi Ren, Jian Wu, Baoye Zhang and Fusheng Zhu
College of Electronics and Information Engineering, Tongji University, 201804
Shanghai, China
GuangDong Communications & Networks Institute, Guangzhou, Guangdong Province, China
e-mail: {1730790, wujun, renhaoqi}@tongji.edu.cn, wujiantongji@126.com, baoye.zhang@tongji.edu.cn,
zhufusheng@gdcni.cn
Abstract—Compared with conventional digital video
transmission, the recently proposed pseudo-analog video
transmission addresses the problem of the cliff effect and
leveling-off effect. However, the researches on pseudo-analog
transmission are mostly done by software, which can’t be
applied in the practice transmission tasks directly. In this
paper, we propose a hardware-optimized pseudo-analog
wireless video transmission system. The transmitted video
signals are divided into the analog part and the digital part,
respectively. The analog part includes the discrete cosine
transformation (DCT) coefficients after power allocation (PA).
The digital part consists of the variance of L-shaped chunks
and some significant fix-point DCT coefficients. Turbo code
and cyclic redundancy check (CRC) are utilized to protect the
digital bits. We propose two optimization methods for the
pseudo-analog system. One of them is to optimize the division
of the source, with which our system performs at most 2dB
better than SoftCast. The other is compression based on
frequency domain data, which achieves a maximum 49%
compression rate with less than 1dB performance loss. We
implement the whole system on FPGA and evaluate the system
with extensive simulations.
Keywords-pseudo-analog; hardware; video transmission;
optimization
I. INTRODUCTION
With the development of multimedia technology, the
demand for wireless transmission grows rapidly. Some real-
time and interactive applications such as First Person View
(FPV) have higher requirements for transmission delay,
while other applications like video broadcasting require
better performance in broadcast scenarios. Due to the limited
error correction capability of the channel coding used in the
digital video transmission system, when the signal-noise
ratio (SNR) is lower than a certain threshold, the quality of
the received video may reach a cliff. That is the cliff effect,
which presents as blocks of mosaic in the video. In the
broadcast scenario, the transmitter of the digital video
transmission system must first predict the exact channel
condition of each receiver to find the worst one. The
transmitter has to send at the lowest bit rate to ensure that
each receiver correctly decodes it. This approach will result
in the receiver with better channel quality not receiving the
video matching its channel quality, which is a waste of
channel bandwidth [1][2].
The pseudo-analog transmission system solves the above
two problems well. The scheme is based on the joint source-
channel coding (JSCC) technology [3][4]. It uses discrete
cosine transformation (DCT) and power allocation (PA) to
replace the quantization in digital transmission system [5].
As a result, pseudo-analog transmission avoids the cliff
effect, making the video quality of each receiver linearly
related to its channel conditions.
SoftCast is a well-known pseudo-analog transmission
system [6][7], which is the basis of other pseudo-analog or
hybrid digital-analog schemes [8][9]. However, most of the
recent researches in the pseudo-analog field are done by
software. As for practical tasks, the hardware-implemented
pseudo-analog transmission system has higher actual value
and better performance. Besides, SoftCast has some limits,
such as occupying a lot of bandwidth, which is also the goal
of optimization.
We propose a pseudo-analog transmission system and
implement it on FPGA. Like other pseudo-analog systems,
the system divides the transmitted data into the digital part
and the analog part [10]. After DCT, power allocation is
utilized to realize unequal protection with the L-shaped
chunk division method [11]. Data after power allocation
constitutes the analog part. During DCT and power
allocation, some low-frequency DCT coefficients and power
of each L-shaped chunk are saved. We use turbo code and
cyclic redundancy check (CRC) to protect it for restoring the
original video better. These digital bits after 16-QAM
modulation constitutes the digital part. We interweave both
of the two parts and put them in the radio frame designed by
ourselves. After sending data via the OFDM channel,
opposite processes are done at the receiver to restore the
origin video.
Besides, we propose two optimization methods for the
pseudo-analog system. One is to improve restoring video
performance. We adjust the division of source by subjoining
the amount of the digital part, which achieves at most 2dB
better performance than SoftCast. The other is to decrease
the demand for the bandwidth of the pseudo-analog system.
We regard the whole 32*32 block as one DC component and
15 L-shaped chunks. The system transmits part of the L-
shaped chunks instead of all of them to decrease the
bandwidth consumption. The experimental results show that
we can reduce half of the data with less than 1dB
performance loss.