FFSCN: Frame Fusion Spectrum Center Net for Carrier Signal Detection

Carrier signal detection is a complicated and essential task in many domains because it demands quick response to the existence of several carriers wideband, while also precisely predicting each carrier signal’s frequency centers bandwidths, including single-carrier multi-carrier modulation signals. Multi-carrier signals, such as FSK OFDM, could be incorrectly recognized signals by using spectrum center net (SCN) or FCN-based method. This paper designed deep convolutional neural network (CNN) framework for fusing features multiple consecutive frames broadband power spectra estimating information broadband, called frame fusion (FFSCN), FFSCN-R, FFSCN-MN, FFSCN-FMN. FFSCN includes three base parts, CNN-based backbone, feature pyramid (FPN) neck, regression (RegNet) head. FFSCN-R FFSCN-MN FPN out features, which use Residual MobileNetV3 respectively, cost less inference time. To further reduce complexity FFSCN-FMN modifies MobileNet blocks fuses at block backbone. The not only preserve high-resolution ratio frequency, but add changes time dimension. Extensive experimental results demonstrate that proposed can effectively detect outperform SCN accuracy efficiency.