| 摘要 |
For the multiplex transmission of coded speech signals in periodic frames, single segments (blocks of coded samples) are selectively suppressed for redundancy reduction, and are replaced on the receiver side by optimally correlated subsections of equal length from previously transmitted segments. On the transmitter side, a multiplicity of compare operations are made for each speech signal, between the respective newest coded segment and a step wise shifted subsection window of previous segments, to determine the best correlated subsection, i.e. the one which is most suitable as replacement and the respective relative offset and correlation. From a group of speech signals, the one signal, or several signals, having the best correlation is selected for suppression of a segment before transmission. Instead of each missing segment an indication of the corresponding optimum offset is transmitted in the frame header. In the receiver for each speech signal a few previous segments are kept in storage from which then, with the aid of the received offset indication the optimally suited subsection can be selected and used as replacement. |
