Audio Watermarking Method Based on the Cochlear Delay Characteristics

This paper proposes a state-of-the-art audio watermarking method, based on the properties of the human cochlear. The method is based on the concept of embedding inaudible watermarks (e.g., a copyright data) into an original sound by controlling the phase characteristics of the sound in relation to the characteristics of cochlear delay. The method consists of two processes: data embedding and data detection. The data embedding process was carried out by applying two types of all-pass filters with the cochlear delays, and then selecting the filtered signal from them according to the watermarks. The data detection process involved by estimating the group delay of the filter from the phase difference between the original and watermarked sounds to detect the embedded data. We designed optimal group delays for the filters. We experimentally evaluated the proposed method by carrying out subjective detection tests and three objective tests of robustness (against down sampling, amplitude compression, and mp3-compression). Results showed that subjects could not detect the embedded data in any of the watermarked sounds we used, and that the proposed method precisely and robustly detected the embedded data from the watermarked sounds.