中文摘要：

如果 watermarked 声音被修改，我们建议能完美地恢复原来的听觉的搬运人的一个新奇听觉的 watermarking 计划。而且，我们能是要求灵活地调整允许的最大的捣弄的率和 watermarked 声音的质量。在计划，有效概括整数变换被改进嵌入在每个补丁，参考小点和检查小点的一般水准的最不重要的位(LSB ) 镇静的水位标数据。LSB 在反的变换被需要。由比较提取检查小点和计算的，当时，修改区域能是局部性的。最后，可靠参考小点和样品数据帮助我们重建没有错误的原来的声音。建议方法的效率理论上并且试验性地被验证。

英文摘要：

We propose a novel audio watermarking scheme which can recover the original audio carrier perfectly if the watermarked audio is modified. Besides, we can adjust the maximum tampered rate allowed and the quality of watermarked audio flexibly as required. In the scheme, an efficient generalized integer transform is improved to embed watermark data which are composed of least significant bits（LSBs） of averages in each patch, reference-bits and check-bits. LSBs are needed in the inverse transform. Then, by comparing the extracted check-bits and calculated ones, the modified area can be localized. Finally, reliable reference-bits and samples data help us reconstruct the original audio without errors. The efficiency of the proposed method is theoretically and experimentally verified.