Real-time full-duplex speech recognition server, based on the Kaldi toolkit and the GStreamer framwork.