Reliable channel data helps characterize the limitations and performance boundaries of communication technologies accurately. However, channel measurement is highly costly and time-consuming, and taking actual measurement as the only channel data source may reduce efficiency because of the constraints of high testing difficulty and limited data volume. Although existing standard channel models can generate channel data, their authenticity and diversity cannot be guaranteed. To address this, we use deep learning methods to learn the attributes of limited measured data and propose a generative model based on generative adversarial networks to rapidly synthesize data. A software simulation platform is also established to verify that the proposed model can generate data that are statistically similar to the measured data while maintaining necessary randomness. The proposed algorithm and platform can be applied to channel data enhancement and serve channel modeling and algorithm evaluation applications with urgent needs for data.

The fifth generation (5G) communication has been a hotspot of research in recent years, and both research institutions and industrial enterprises put a lot of interests in 5G communications at some new frequency bands. In this paper, we investigate the radio channels of 5G systems below 6 GHz according to the 5G communication requirements and scenarios. Channel measurements were conducted on the campus of Beijing Jiaotong University, China at two key optional frequency bands below 6 GHz. By using the measured data, we analyzed key channel parameters at 460 MHz and 3.5 GHz, such as power delay profile, path loss exponent, shadow fading, and delay spread. The results are helpful for the 5G communication system design.