Blind mode detection for AMR speech data

In this paper, we propose a blind method to detect the mode of AMR speech data. This method provides a compression mode information by use of investing bit stream of unknown compressed data. Unlike uncompressed data, compressed data is inevitable in communication channels due to their efficient data size. However, the compressed data can lose their critical information such as the mode of AMR since the information is possibly delivered in a different channel or lost in a memory device. For these reasons, we propose an algorithm to detect the mode information of AMR when the mode information is missing. The algorithm consists of the processes of analyzing the compressed bitstreams and defining a specific group of bits to detect the mode information efficiently. To evaluate our method, we conduct experiments. Our experiments indicate that the proposed method provides sufficient performance to classify the mode information of AMR.