Method and system for packetised content streaming optimisation

摘要

A method of determining the speech content of a packet carrying speech encoded data missing from speech segment communicated by in a packetized data stream communicated using at least one VOIP link between a server platform and a client platform, the method comprising at the client platform: receiving a plurality of packets carrying speech encoded data forming said packetized data stream; processing each received packet to determine a unique message segment identifier associated with a speech segment of the received packet; processing each received packet to determine if it contains another unique message segment identifier associated with a previously received packet carrying encoded speech data; determining if the unique message segment identifier for the received packet exists in storage means provided on the client platform, and if not, storing the received packet in association with its unique message segment identifier; processing each received packet to determine a sequence identifier; checking if the sequence identifier is contiguous in sequence with a previously received packet stored locally on said client platform, and if not, determining the speech content of one or more missing packet in the sequence sent by the server platform to the client platform by retrieving a packet from said storage means having the same unique message segment identifier as the missing packet.

主权项

1. A method of maintaining the quality of speech recognition in an interactive voice recognition system comprising at least one client platform arranged to communicate over at least one communications link, capable of supporting a connectionless data packet connection, with an interactive voice recognition server platform, the method comprising the steps of:
predicting a grammar complexity for a future speech segment to be received in an encoded form from said client platform responsive to a prompt provided by an IVR application running on said server platform; and determining if the encoding to be used in said future speech segment to be received by said server platform can be optimised if said prediction indicates a change in grammar complexity, and if so:
generating based on the predicted grammar complexity of said future speech segment a new codec to be used by the client platform for encoding of said future speech segment;transmitting said new codec to said client platform in at least one packet forming a data stream arranged to generate said prompt at said client platform; andcommanding said client platform to encode subsequent speech segments with said new codec.