Specific Emitter Identification (SEI) is a critical technology used to uniquely identify radio emitters based on their Radio Frequency Fingerprint (RFF) extracted from received signals. Deep learning (DL), which has proven effective in many recognition tasks, is believed to benefit SEI by strengthening wireless security authentication through RFFs. In this paper, we present a new approach for emitter identification that employs a feature late fusion strategy. Specifically, in order to identify eight LoRa transceivers of same model, a residual neural network-based DL model is presented which integrates the short-time Fourier transform (STFT) spectrograms with statistical features extracted from the in-phase and quadrature (IQ) signals, achieving a fusion of these two feature sets. The experimental results show that the proposed fusion strategy significantly improves recognition accuracy compared to using either STFT spectrograms or statistical features alone.