Digital audio is increasingly becoming more and more a part of our daily lives. Unfortunately, the excessive bitrate associated with the raw digital signal makes it an extremely expensive representation. Applications such as digital audio broadcasting, high definition television, and internet audio, require high quality audio at low bitrates. The field of audio coding addresses this important issue of reducing the bitrate of digital audio, while maintaining a high perceptual quality. Developing an efficient audio coder requires a detailed analysis of the audio signals themselves. It is important to find a representation that can concisely model any general audio signal. In this thesis, we propose two new high quality audio coders based on two different audio representations - the sinusoidal-wavelet representation, and the warped linear predictive coding (WLPC)-wavelet representation. In addition to high quality coding, it is also important for audio coders to be flexible in their application. With the increasing popularity of internet audio, it is advantageous for audio coders to address issues related to real-time audio delivery. The issue of bitstream scalability has been targeted in this thesis, and therefore, a third audio coder capable of bitstream scalability is also proposed. The performance of each of the proposed coders was evaluated by comparisons with the MPEG layer III coder. The first coder proposed is based on a hybrid sinusoidal-wavelet representation. This assumes that each frame of audio can be modelled as a sum of sinusoids plus a noisy residual. The discrete wavelet transform (DWT) is used to decompose the residual into subbands that approximate the critical bands of human hearing. A perceptually derived bit allocation algorithm is then used to minimise the audible distortions introduced from quantising the DWT coefficients. Listening tests showed that the coder delivers near-transparent quality for a range of critical audio signals at G4 kbps. It also outperforms the MPEG layer III coder operating at this same bitrate. This coder, however, is only useful for high quality coding, and is difficult to scale to operate at lower rates. The second coder proposed is based on a hybrid WLPC-wavelet representation. In this approach, the spectrum of the audio signal is estimated by an all pole filter using warped linear prediction (WLP). WLP operates on a warped frequency domain, where the resolution can be adjusted to approximate that of the human auditory system. This makes the inherent noise shaping of the synthesis filter even more suited to audio coding. The excitation to this filter is transformed using the DWT and perceptually encoded. Listening tests showed that near-transparent coding is achieved at G4 kbps. The coder was also found to be slightly superior to the MPEG layer III coder operating at this same bitrate. The third proposed coder is similar to the previous WLPC-wavelet coder, but modified to achieve bitstream scalability. A noise model for high frequency components is included to keep the overall bitrate low, and a two stage quantisation scheme for the DWT coefficients is implemented. The first stage uses fixed rate scalar and vector quantisation to provide a coarse approximation of the coefficients. This allows for low bitrate, low quality versions of the input signal to be embedded in the overall bitstream. The second stage of quantisation adds detail to the coefficients, and hence, enhances the quality of the output signal. Listening tests showed that signal quality gracefully improves as the bitrate increases from 16 kbps to SO kbps. This coder has a performance that is comparable to the MPEG layer III coder operating at a similar (but fixed) bitrate.

For this study three questions are being studied. Question one: “How does the audio compressionaffect the perceived video quality?” Question two is somewhat reversed: “How does the texturedetail affect the perceived audio quality?” The final question follows: “How do the audiocompression and texture detail affect the perceived overall quality?” The experiment was done on 36different untrained listeners, whereof 2 didn’t give complete answers and were therefore notincluded in the results. In total the result was based on 34 test subjects. Subjects participating in theexperiment had to play through three levels, one level for each research question. In each levelsubjects evaluated the game’s different qualities. In one level only the texture detail changed (LevelVV), in the other only audio compression rate changed (Level AA) and in the third both audiocompression and texture detail changed (Level AV). The texture detail ranged from low to medium tohigh setting, while the audio compression had four different levels ranging from 49 kbit/s to 150kbit/s, in the compressed format of ogg vorbis. The result shows that subjects did not perceive animprovement in quality in either of the single quality tests, i.e. levels AA and VV. In the last level withmultiple changing stimuli (AV), subjects could identify the lower video quality from medium and high.Subjects also showed a significant difference in perceived overall quality in the AV level, subjectsshowed that they could identify a difference in audio quality better, when video quality was on thelow setting.

The increasing interest for immersive experiences in areas such as augmented and virtual reality makes high quality 3D sound more important than ever before. A technique for capturing and rendering 3D audio which has received more attention during the last twenty years are Higher Order Ambisonics (HOA). Higher Order Ambisonics is a scene based audio format which has a lot of advantages compared to other standard formats. Hovever, one problem with HOA is that it requires a lot of bandwidth. For example, sending an uncoded high quality HOA signal requires 49 channels to be transmitted at the same time which requires a bandwidth of about 40 Mbps. A lot of effort has been made in the last ten years on coding HOA signals. In this thesis, two different approaches are taken on coding HOA signals. In one approach, called Sound Field Rotation (SFR) in this thesis, the microphone that records the sound field is virtually rotated to see if it is possible to make some of the channels zero. The second approach, called Sound Field Decomposition (SFD) in this thesis, use Principal component analysis to decompose a sound field into a foreground and background component. The Sound Field Decomposition approach is inspired by the emerging MPEG-H 3D Audio standard for coding HOA signals. The result shows that the Sound Field Rotation method only works for very simple sound scenes. It has also been shown that a 49 channels HOA signal can be reduced to as little as 7 channels if the sound scene consists of a point source. The Sound Field Deomposition method worked for more complex sound scenes. It was shown that a MPEG similar system could be improved. Result from MUSHRA (Multiple stimuli with hidden reference and anchor) listening tests showed that an improved MPEG similar system reached a MUSHRA score about 78 while the MPEG similar system reached 55 at a bitrate of 256 kbps. Without coding each monochannels with the 3GPP EVS (Enhanced voice services) codec, the improved MPEG similar system reached the MUSHRA score 85. At 256 kbps, the improved MPEG similar system coded the HOA signal into six channels instead of 49 for the uncoded signal. From objective results, it was shown that the improved MPEG similar system had largest effect at low bitrates.

Submitted by William Justo Figueiro (williamjf) on 2015-07-08T20:56:12Z No. of bitstreams: 1 03b.pdf: 3126214 bytes, checksum: 0f98dbf86ae74816af91944aa7dec80f (MD5) / Made available in DSpace on 2015-07-08T20:56:12Z (GMT). No. of bitstreams: 1 03b.pdf: 3126214 bytes, checksum: 0f98dbf86ae74816af91944aa7dec80f (MD5) Previous issue date: 2013 / Nenhuma / A qualidade perceptual alcançada pelos codificadores de áudio depende diretamente da escolha de seus parâmetros. O codificador MPEG-4 AAC (Advanced Audio Coding), utilizado no Sistema Brasileiro de Televisão Digital (SBTVD), possui em sua estrutura uma etapa composta por um laço de iteração para escolher os parâmetros do codificador, de maneira dinâmica durante o processo de codificação. Este processo de escolha pode ser definido como um problema de Pesquisa Operacional, sendo um problema de Seleção de Partes, denominado como o Problema de Codificação AAC. A estrutura existente no codificador de referência, não resolve este problema de maneira ótima. Desta forma, este trabalho propõe o desenvolvimento e implementação de um modelo de uma estrutura de simulação, para encontrar os parâmetros do codificador de áudio MPEG-4 AAC, de maneira a otimizar a qualidade perceptual do áudio, para uma determinada taxa de bits (bit rate). A implementação da estrutura de otimização foi desenvolvida em linguagem C, utilizando as metaheurísticas Busca Tabu e Algoritmo Genético em uma estrutura híbrida. Através da minimização da métrica ANMR (Average Noise-to-Mask Ratio), o algoritmo procura identificar a melhor configuração dos parâmetros internos do codificador MPEG-4 AAC, de maneira que possa garantir uma qualidade perceptual para o sinal áudio. Os resultados obtidos utilizando a estrutura híbrida de otimização apresentaram valores menores para a métrica ANMR, ou seja, uma melhor qualidade perceptual de áudio, quando comparados com os resultados obtidos com o codificador de referência MPEG-4 AAC. / The perceptual quality achieved by audio encoders depends directly on the choice of its parameters. The MPEG-4 AAC (Advanced Audio Coding), used in the Brazilian Digital Television System (BDTS), has a step in its structure that consists in iteration loop to choose the parameters of the encoder dynamically during the encoding process. This selection process can be defined as a problem of Operational Research, being a Part Selection Problem, termed as AAC Encoding Problem. The structure in the reference encoder not solves this problem optimally. Thus, this paper proposes the development and implementation of a model simulation of a structure, to find the internal parameters of the MPEG-4 AAC audio encoder, so as to optimize the perceptual audio quality for a given bit rate. The implementation of the optimization framework was developed in ANSI C programming language, using the Tabu Search and Genetic Algorithm metaheuristics in a hybrid structure. Through the minimization of the ANMR (Average Noise-to-Mask Ratio) metric, the algorithm tries to identify the best configuration of internal parameters of the MPEG-4 AAC. The results obtained using the optimization hybrid structure achieve lower values for the ANMR metric, i.e., an better perceptual audio quality, compared with the obtained with the reference encoder MPEG-4 AAC.

<p dir="ltr">The rapid development of machine learning techniques makes it possible to manipulate or synthesize video and audio information while introducing nearly indetectable artifacts. Most media forensics methods analyze the high-level data (e.g., pixels from videos, temporal signals from audios) decoded from compressed media data. Since media manipulation or synthesis methods usually aim to improve the quality of such high-level data directly, acquiring forensic evidence from these data has become increasingly challenging. In this work, we focus on media forensics techniques using the metadata in media formats, which includes container metadata and coding parameters in the encoded bitstream. Since many media manipulation and synthesis methods do not attempt to hide metadata traces, it is possible to use them for forensics tasks. First, we present a video forensics technique using metadata embedded in MP4/MOV video containers. Our proposed method achieved high performance in video manipulation detection, source device attribution, social media attribution, and manipulation tool identification on publicly available datasets. Second, we present a transformer neural network based MP3 audio forensics technique using low-level codec information. Our proposed method can localize multiple compressed segments in MP3 files. The localization accuracy of our proposed method is higher compared to other methods. Third, we present an H.264-based video device matching method. This method can determine if the two video sequences are captured by the same device even if the method has never encountered the device. Our proposed method achieved good performance in a three-fold cross validation scheme on a publicly available video forensics dataset containing 35 devices. Fourth, we present a Graph Neural Network (GNN) based approach for the analysis of MP4/MOV metadata trees. The proposed method is trained using Self-Supervised Learning (SSL), which increased the robustness of the proposed method and makes it capable of handling missing/unseen data. Fifth, we present an efficient approach to compute the spectrogram feature with MP3 compressed audio signals. The proposed approach decreases the complexity of speech feature computation by ~77.6% and saves ~37.87% of MP3 decoding time. The resulting spectrogram features lead to higher synthetic speech detection performance.</p>