Decoder synchronization adjustment
View Patent ↗A device is disclosed that makes packetized and encoded speech data audible to a listener, as is a method for operating the device. The device includes a unit for generating a synchronization request for reducing an amount of synchronization delay, and further includes a speech decoder that is responsive to the synchronization delay adjustment request for executing a time-warping operation for one of lengthening or shortening a duration of a speech frame. In one embodiment the speech decoder comprises a code excited linear prediction (CELP) speech decoder, and the CELP decoder time-warping operation is applied to a reconstructed excitation signal u(k) to derive a time-warped reconstructed signal u w (k). The time-warped reconstructed signal u w (k) is input to a Linear Predictor (LP) synthesis filter to derive a CELP decoder time-warped output signal y^ w (k) In another embodiment the speech decoder comprises a parametric speech decoder, and where an adaptation of the frame length N in the parametric speech decoder results in the use of a modified frame length N w .
1. A mobile telephone comprising:
a receiver configured to receive encoded audio data comprising packetized VoIP speech data;
a jitter buffer controller configured to generate a synchronization request in response to a determination based at least in part on an amount of time that audio data resides in a jitter buffer to make the average duration a frame resides in the jitter buffer to approach a desired jitter buffer residency duration; and
a decoder of the received encoded audio data, the decoder operating in response to the synchronization request to execute a time-warping operation to lengthen a duration of a speech frame,
where the decoder comprises a code excited linear prediction decoder, and where a code excited linear prediction decoder time-warping operation is applied to a reconstructed excitation signal u(k) to derive a time-warped reconstructed signal u w (k), and where the time-warped reconstructed signal u w (k) is input to a linear predictor synthesis filter to derive a code excited linear prediction decoder time-warped output signal y^ w (k).