Kotaro Sonoda

Audible Secret Keying for Time-Spread-Echo Based Audio Watermarking

This paper deals with the pseudo noise (PN) generating method for digital audio watermarking using time-spread echo hiding. In time-spread echo based audio watermarking, the secret payload is embedded in the form of multiple echoes spread by a pseudo noise sequence and the pseudo noise sequence is used as secret key. Generally, the pseudo noise sequence is required to be uncorrelated with other sequence and therefore typically generated by an M-sequence generator. In this paper, we propose a key sequence generating method which generates a key sequence from an audio signal. By using the new key sequence generated from an audio signal instead of an M-sequence, the key information users have to remember is their secret audio signal but not a complicated random PN key. Therefore, the availability of digital audio watermarking would be improved. After introducing the new key generation method, we evaluated the proposed key sequence. The result shows it can detect a secret payload similar to use of a conventional M-sequence and does not deteriorate inaudibility of the embedded watermark.

Audio Watermarking Using QIM on Wavelet Packet Coefficients

In this paper, we propose an audio watermarking method that is based on wavelet packet transform (DWPT) coefficients. One bit of watermark codes is embedded into the power ratio (quantization index) of a pair of DWPT coefficients by quantization index modulation (QIM). The encoded pair is selected from the higher harmonic regions of the most powerful coefficient in a DWPT. By using such a criterion to select the watermarked region, we can expect to deal with the conflict of the length of analyzing between the sides of embedding watermarks and detecting watermarks. Multiple bits are embedded into the several pairs that are selected from the coefficient series of a DWPT. If the selected pair was not expected to detect the power ratio, the watermark bit is not embedded into such a pair. This realizes various bit rate embedding of watermark.

Information Hiding for Public Address Audio Signal using FH/FSK Spread-spectrum Scheme

We present a novel application and technology for hiding the messages into the emergency public address audio signals. In the emergent situation such as disaster, the public address siren is very effective way to notify the urgency of disaster to residents and supporting staffs. The people can estimate the direction of the disaster spot or distance from the spot by hearing and seeing the siren cars moving on the way. However, the people in house or vehicle often cannot hear the notifications. Our study helps them to receive the notifications embedded into the public address siren such a scene. Moreover, the supporting staffs, who work outside, can receive the information about the helps from the siren on the way. To realize our aims, high resistancy of data transmission under outside noisy environment and high acceptance to receive from the moving signal sources are necessary. In this paper, we propose a public address information hiding (PAIH) using FH/FSK spread-spectrum scheme and evaluate its performance with the computer simulations.

Phase rotation thresholds and its use in watermarking technique

Strong demands for the protection of copyrights for high fidelity audio recordings as well as broadcasting signals via a network brought about a development of so‐called watermarking techniques. One of them is based on a phase modulation referred to in here as a phase rotation (PR). This method assumes a poor sensitivity of the auditory system to the phase changes in a signal and embeds a watermark in the recording by means of a filter with linear phase characteristics. However, the sensitivity of the auditory system to the phase rotation has not been studied in detail, especially for complex signals. In this paper we present a preliminary result of threshold measurements of the phase rotation applied to different types of signals. It appears that the thresholds for detecting the phase rotation, when expressed by means of a time group delay, depend on the type of the signal and the frequency region that the phase rotation is applied for. The detection of the phase rotation is approximately independent on ...

Tally Based Digital Audio Watermarking

In this paper, we propose a novel digital audio watermarking system inspired by the tally trade. In ordinary digital audio watermarking, a single stego signal is produced through the embedding process and the hidden message is extracted from the stego signal. In tally-based system we propose, multiple stego signals are tallied to produce and the hidden message which is extracted from the temporally mixed signal composed of the required number of stego tallies. When the extractor can’t mix the required number of stego tallies, it misses the hidden message. The system’s performances are evaluated in terms of the number of tallies (shares).

Digital audio watermarking robust against Locality Sensitive Hashing

Audio fingerprinting is an audio feature vector and a technique to identify the music source by comparing with original fingerprints in a fingerprint database. The fingerprints are often hashed to shorten codes. The content similarities between fingerprints are maintained in the hashed code. Such a hash function family is called Locality Sensitive Hashing (LSH). Because audio fingerprint using LSH is known to be resistant against CPO (Content Preserving Operations, perceptively acceptable manipulations) such as compression, noise adding, mean filtering, it is possible to identify the original source even if the source was slightly modified. On the other hand, mixed arrangement (mashup) of several music sources is allowed as legitimate artistic expression. In the conventional fingerprint based retrieval system, the mixed arrangements could identify the origins segmentally but the arranger’s authorization is ignored. In this report, we propose an audio watermarking method robust against LSH coding. That is, the arranger information is watermarked in the audio signal and it is detectable from not only stego audio signal but also stego audio fingerprint of LSH.

Digital Watermarking Method Based on STFT Histogram

This paper presents a digital audio watermarking method. In the method, watermark has been represented by a form of the histogram for sound power levels distribution in time domain in certain frequency band(s). The relation of four numbers of selected bins in a histogram are changed by a rule of watermark bits assignment. To evaluate the performance of the embedding procedure in each frequency band, the robustness test against several kinds of manipulations or attacks were conducted for some instrumental music sources and harmonic complex tone. Considering the results of robustness test, the performance of multi-band embedding is discussed. Promising results of hearing tests are also presented.

User Authentication Scheme Using Individual Auditory Pop-Out

This paper presents a user authentication scheme which takes advantage of the uniqueness of prover’s auditory characteristics. In this scheme, several audio stimuli are presented at the same time by headphone. The stimuli include a special audio stimulus that only the genuine prover can easily distinguishes from other stimuli because of the uniqueness of auditory characteristics. The prover is made to answer the contents of the special stimulus. The verifier confirms the correct answer as the genuine prover. As the special audio stimulus that distinguishes the genuiness, auto-phonic production is examined in this paper.

Audio CoFIP (contents fingerprinting) robust against collusion attack

Digital fingerprinting method allows individualization of all content purchased by users. In a multi-cast network, however, distributed contents must be identical among all users. Therefore, the process of individualizing content should be given on the user side while the original signal should not be delivered to users for security reasons related to copyright protection. Content fingerprinting (CoFIP) for audio signals was proposed to meet both requirements. In CoFIP, a problem might occur when users collude. Synchronous addition of several individualized audio signals makes it difficult to identify each individualized contents composed in it. This article proposes a countermeasure against such collusion, providing imperceptible phase shift in the individualization. Listening tests and computer simulations show that the fingerprint can be detected correctly while the sound quality of the colluded signal is effectively degraded.