Page 187 - Kaleidoscope Academic Conference Proceedings 2021

WIRELESS CHANNEL SCENARIO RECOGNITION BASED ON NEURAL NETWORKS

1, 2
1, 2
Xiaojing Xu ; Ruimei Li ; Hua Rui ; Wei Lin ; Xiangfeng Liu ; Wei Cao
2
1, 2
1, 2
1, 2
State Key Laboratory of Mobile Network and Mobile Multimedia Technology , ZTE Corporation , China
2
1
ABSTRACT recognition, etc. This paper mainly focuses on the research
work of combining machine learning and wireless channel
Wireless channel scenario recognition plays a key role in scenario recognition.
improving the performance of mobile communication
systems. This paper combines wireless channel In [6], a machine learning method is proposed for indoor
characteristics extracted using expert experience and environment classification based on real-time measurement
neural networks, and proposes a wireless channel scenario of radio frequency signals, which mainly distinguishes the
recognition framework based on neural networks. Firstly, four scenarios of indoor high, medium, and low scattering
the wireless propagation environment is analyzed, and and open space. The statistical characteristics of Received
some wireless channel characteristics are extracted, such Signal Strength Indication (RSSI) values are used in [7] to
as the frequency domain fading factor, multipath power distinguish indoor Line-Of-Sight (LOS)/Non-Line-Of-Sight
delay distribution, time domain energy peak response ratio (NLOS) channel scenarios. In [8], a new Convolutional
and time correlation characteristics. Secondly, the Neural Network (CNN) algorithm for ultra-wideband
combined algorithm model using the wireless channel channel classification is proposed, which is aiming at
characteristics and neural networks are proposed. Finally, indoor office LOS/NLOS. In [9], a LOS/NLOS channel
after simulation verification, the new method has a greater identification method based on machine learning is also
improvement in the recognition accuracy than the proposed for ultra-wideband indoor positioning systems.
traditional threshold algorithm. The combination of angle information and support vector
machines can significantly improve the accuracy of
Keywords – Channel characteristics, neural network, LOS/NLOS channel scene identification in [10]. A machine
scenario recognition, wireless channel learning-based intelligent vehicle communication scene
recognition model is proposed in [11], and the results show
1. INTRODUCTION that the recognition accuracy rate is above 98% in four
typical scenarios, including urban areas, highways, tunnels,
In a wireless communication system, the wireless channel and vehicle obstacles. A machine learning technology for
has a huge impact on the entire system. Due to the complex LOS/NLOS recognition in Internet of Vehicles is developed
and changeable nature of the channel environment, the in [12], which mainly used the power angle of the channel
signals under different channel scenarios have obvious impulse response as channel characteristics for training and
differences in signal energy, time delay and channel recognition. The distinction of high-speed railway wireless
response. If different channel scenarios can be accurately channel scenarios is studied in [13] and [14], and is also
identified, the receiver can adopt the optimal algorithm and used machine learning or deep learning. The power spectral
configuration parameters to match the channel. So, the density of the channel autocorrelation function is extracted
wireless channel scenario recognition is of great in [15], and is combined with the deep belief network to
significance to improving the performance of the wireless identify the wireless channel scenarios.
communication system.
The current research mainly focuses on the indoor channel
As the most popular field in recent years, machine learning scenario, Internet of Vehicles channel scenario and high-
th
has been widely used in data mining, pattern recognition, speed railway channel scenario. In 4 -Generation (4G) or
th
image processing, natural language processing, unmanned 5 -Generation (5G) commercial mobile communication
driving, etc. The combination of machine learning and systems, dense urban scenarios and rural or suburban
wireless communication has also attracted the attention of scenarios are also the main communication scenarios. If the
many scholars, and there have been many attempts, such as different channel scenarios in these communication
Artificial Intelligence (AI)-based modulation recognition [1] scenarios can be identified, the receiving algorithm can
[2], AI-based channel estimation and detection [3], AI- make adaptive adjustments based on the identified wireless
based channel coding and decoding [4], AI-based channel channels to improve the performance of 4G or 5G
modeling [5] and AI-based wireless channel scenario communication systems. For example, the receiver can
select a filter with an appropriate window length for

978-92-61-33881-7/CFP2168P @ ITU 2021 – 125 – Kaleidoscope