Deepfake Audio Detection Using Spectrogram-based Feature and Ensemble of Deep Learning Models