media resource control protocol (MRCP)
The Media Resource Control Protocol (MRCP) allows speech servers to provide various speech services to their clients for speech recognition, speech processing, or speech synthesis.
For this purpose, the MRCP protocol works with requests and responses. To do this, MRCP clients send requests to the server. These requests, which, like all communication, are transmitted via transport protocols such as the Realtime Streaming Protocol( RTSP) or the TCP protocol, may include, for example, whether the MRCP client can send speech data to the MRCP server for speech processing.
MRCP is similar in structure to other Internet protocols such as HTTP and consists of three packet parts. The first packet part contains the message type and other information such as the code for the response. The second packet part is the header, which contains the number of lines and the scope of the message part, and the third packet part is the body with the details of the messages.
MRCP can work with streaming protocols such as RTSP or the Session Initiation Protocol( SIP), which can be used to connect to external streaming devices and streaming mechanisms such as the RTP protocol.
MRCP was developed by Cisco, Nuance, and Speechworks and is under development by the Internet Engineering Task Force( IETF). It is described in RFC 4463.