Silent Speech Interfaces (SSIs) have a broad range of applications and rely on sensing techniques to measure speech-related bio-signals. To take advantage of the strengths and minimize weaknesses of different sensing techniques, multimodal recordings and pipelines have been investigated lately on SSIs. This study presents a multimodal articulatory recording combining Radar and Optopalatographic (OPG) SSIs and evaluates the performance of each data type and their combination in a phoneme recognition task. A corpus of 26 phonemes was recorded by 3 native German speakers and Support Vector Machines were used to classify the recorded multimodal data into the phonemes. Considering all speakers together, accuracies of 85.40%, 66.99% and 86.55% were obtained with Radar, OPG and the concatenation of Radar and OPG, respectively. The accuracy increase of 1.15% resulting of adding OPG to Radar data is small but statistically significant and points to the potential of such multimodal recordings.