Improved packets loss concealment in speech coding by data hiding

摘要

Effectiveness of Code Excited Linear Prediction speech codecs dwells in a good reconstruction of excitation related to the mixture of both: adaptive codebook and fixed codebook excitations. Therefore, in this study we focused our attention on the lost frame excitation reconstruction by using data hiding. Hence, the goal behind this idea is to add important information about excitation for use in case of an eventual loss, but without any overload of the transmission bandwidth to avoid increasing the packets congestion in the network. Indeed, we have to find a good trade-off between the embedded data and keeping the intelligibility of speech after decoding as well as possible. In our study, we appeal to data hiding technique to hide samples about previous excitation sent frame in the recent excitation samples frame in process. Therefore, the hidden data related to any frame can be extracted once the following frame, coming just after, will be received; and this with respect to the considered delay between two successive frames. The test results show that this method is a promising area for future works on frame loss recovery using data hiding.