Radio Frequency (RF) fingerprinting is an emerging technology for advanced device authentication. In this work, we investigate the feasibility of utilizing three different types of fundamental waveform for the purpose of RF fingerprinting. The Short-time Fourier Transform (STFT) is adopted to exploit potential RF fingerprints, which results are then converted into spectrograms and are used to train a convolutional neural network (CNN) for device authentication. An experimental setup constructed using five software-defined radios (SDRs) is tested against an additive white Gaussian noise (AWGN) channel to verify its performance under varying signal-to-noise ratios (SNRs). The experimental results show a promising classification accuracy ranging between 94.1% and 99.2% depending on the utilized waveform.