Audio streaming's Project comes to a paragraph, summarizing the key technologies: 1. Improvement of the MP3 encoder 1) Remove Interleave to the HUFFMAN CODE section 2) Separate critical, scf, and Huffman, different protection according to the importance 3) Different protection 3) Hierarchical processing, divided into 2 part according to Perceptual Effect 4) Bit Reserver technology is taken to the Critical section, so that Cirtical's Error Propagatiojn may further reduce 2. Error ConceAlMent 1 of the MP3 decoder is more than the effect of gravity! 2) Surprisingly 3 by using the upper frame of SCF recovery loss of Frame (simple COPY) effect. Network adaptation 1) retransmit, returning to the critical section, Reason is small data, only about 4%, but is critical. Moreover, the improved MP3 determines at least 64 Frame delay, which can be fully utilized 2) Forward Error Correction: Simple XOR, and Duplication Duplication is very poor, and during Dupliation The original frame is again used by Interleave, the loss of two packages can only cause 1 Frame loss 3) Enhance Layer's huffmancode can be implemented by Discard: 1. Environment: WinXP, Centrino Adhoc Wi-Fi 2. Receive all the working threads and sockets and reduce the burden. 3. After the improvement, MP3 needs 60 Frame simultaneously decoded. The time CPU occupies greatly, it may affect the Socket (Code Optimization Problem Solution). If you encounter similar situations in the future, you can divide a decoding process into several sections and put it in different times. 4. The Timer of BCB is too incorrect, but this project has little impact, and then rewriting the MultiMedia-specific. . . 5. Error handling, including buffer overflow, decoding fail, player is expected to be in this case. . . otherwise. . . Question lessons: 1. Adhoc's packet loss is not a proportion of distance, but it suddenly decreases countless after a little. . . Resulting in data that cannot be acquired 2. The packet loss situation in the same place is extremely different at different times, although the program can cope with this mutation, but this data is not suitable for testing. Finally, I have to simulate with Error Pattern File, SIGN 3. The template does not use, the result of the template is: I don't feel it for the caller, but the implementation of the programming work will not be reduced, and it has increased. Only inheritance can reduce the amount of code. (Unless STL ...) 4 is intended.
