초록 ▼

To reduce the needed bandwidth of video communication system, a portion of the video image that differs from a preceding frame of the video image is predicted from the decompressed audio data with a model operating at both the transmitter and the receiver. The model accuracy is increased and the syn

To reduce the needed bandwidth of video communication system, a portion of the video image that differs from a preceding frame of the video image is predicted from the decompressed audio data with a model operating at both the transmitter and the receiver. The model accuracy is increased and the synchronisation enhanced by reducing the number of degrees of freedom of the model and by making the prediction of the next phoneme in a hierarchical manner.

대표청구항 ▼

1. A method of operating a video communication system comprising; monitoring an event with a video camera to generate a sequence of frames for forming a video image; selecting video data only from those regions of a current frame of the video image that are different from corresponding regions o

1. A method of operating a video communication system comprising; monitoring an event with a video camera to generate a sequence of frames for forming a video image; selecting video data only from those regions of a current frame of the video image that are different from corresponding regions of a preceding frame of the video image; comparing one or more of said selected regions of video data to stored video regions of a transmitter facial model; removing from said selected regions any regions found to compare within a defined limit to regions stored in said facial model; compressing video data corresponding to said selected regions of the current frame of the video image (less those regions removed in the previous step) and audio data for each of said frames of said video image; generating a video signal comprising compressed video data and compressed audio data and transmitting said video signal to a receiver; receiving said transmitted video signal at a receiver; decompressing said received video signal to produce audio data and a first portion of said video data; predicting a second portion of the video data for regions of a current frame of a video image that differ from a preceding frame of the video image from said audio data by adding back into the video data data determined by a receiver facial model; and combining said first and second portions of said video data to generate the current frame of a display video image; wherein said facial model predicts an eigen phase based on a previously uttered sub-phoneme. 2. The method as claimed in claim 1 wherein prediction of an eigen phase based on a previously uttered sub-phoneme is done at both the transmitter and receiver facial models. 3. The method as claimed in claim 1 wherein prediction of an eigen phase based on a previously uttered sub-phoneme is done with a neural network. 4. A video communication system comprising; a video camera for monitoring an event and for generating a sequence of frames for forming a video image; means for selecting video data only from those regions of a current frame of the video image that are different from corresponding regions of a preceding frame of the video image; a data compression module for compressing video data corresponding to said selected regions of the current frame of the video image and audio data for each of said frames of said video image; means for generating a video signal comprising said compressed video data and compressed audio data and transmitting said video signal to a receiver; a data decompression module for decompressing a received video signal to produce audio data and a first portion of said video data; means for predicting a second portion of the video data for regions of a subsequent frame of the video image that differ from a preceding frame of the video image from said audio data based on a sub-phoneme; and means for combining said first and second portions of said video data to generate a display image.