This paper describes a new method for music onset detection. The novelty of the approach consists mainly of two elements: the time-frequency processing and the detection stages. The resonator time frequency image (RTFI) is the basic time-frequency analysis tool. The time-frequency processing part is in charge of transforming the RTFI energy spectrum into more natural energy- change and pitch-change cues that are then used as input elements for the detection of music onsets by detection tools. Two detection algorithms have been developed: an energy-based algorithm and a pitch-based one. The energy-based detection algorithm exploits energy-change cues and performs particularly well for the detection of hard onsets. The pitch-based algorithm successfully exploits stable pitch cues for the onset detection in polyphonic music, and achieves much better performances than the energy-based algorithm when applied to the detection of soft onsets. Results for both the energy-based and pitch-based detection algorithms have been obtained on a large music dataset.

The automatic analysis of a polyphonic sound is still a very challenging task, not only for computational reasons but also because of the lack in suitable techniques and restrictions in the application field, or sometimes due to unrealistic goals. A remarkable progress has been made in the last decade, but still, a practical and generic solution for this problem is hard to find. In this paper, we propose a rather general solution for a chord/harmony analyzer, which is able provide good results for different instruments and polyphonic sounds. It is based on the combination of signal processing and neural networks. The sound is analyzed in both time and frequency domains by a hybrid original technique. A rather innovative approach is then used to classify the chords and extract their evolution time. The proposed overall method aims to implement a general purpose listening machine, whose approximated results and approach are nevertheless general enough to allow the implementation of very useful applications.

The MPEG-4 audio standard provides several toolsets for natural and synthetic sound coding. Among them, the most innovative in terms of multimedia applications is Structured Audio (SA), which implements a high level, structured description of sound instead of the usual compression techniques based on psychoacoustics and subband analysis. SA permits to encode synthesis and processing algorithms by its Structured Audio Orchestra Language (SAOL), and it can theoretically be used to specify any other audio decoder. This great flexibility introduces a challenging implementation problem, which in a normative framework has to be solved by a systematic approach. In the first part of the paper, it is described how the SA decoding process can be analyzed in a platform independent way in order to determine the fundamental figures of this coding technique; it is then shown how the proposed method is being used for the MPEG-4 SA conformance test. In the second part the design of a virtual digital signal processor (DSP) architecture is presented, based on the results of the complexity analysis. This architecture is able to exploit the intrinsic data level parallelism of many Audio algorithms and to consistently reduce the implementation cost. Experimental results prove the effectiveness of the approach and its suitability for implementations on modern superscalar DSPs and multimedia processors.

Natural and structured audio representations can be characterized by the lack or presence of a model describing the sound, respectively; combination of the two approaches can lead to efficient and improved storage and transmission of both speech and music, mixing less efficient but general technologies with more compact and specialized models. Integration of natural audio tracks with structured sound and 3D spatial processing is a challenging effort, especially when the audio, scene requires high quality and precise synchronization with video and graphic information, as it is the case in professional multimedia and virtual reality frameworks. In this paper natural and structured sound are surveyed and a new player is presented, which supports all the mentioned technologies in a normative context.

The new MPEG-4 audio standard provides two toolsets for synthetic audio generation, audio processing and multimedia content description called Structured Audio (SA) and Binary Format for Scenes (BIFS). Moving from a systematic analysis of SA and from the implementation of an efficient SA decoder, this paper describes the design of a virtual DSP architecture able to exploit the data level parallelism contained in many typical audio processing algorithms. The proposed virtual DSP architecture shows good performance on general purpose platforms and can be easily adapted and optimized for parallel superscalar devices. The porting and results on a V-LIW DSP device confirm the effectiveness and flexibility of the approach, which is particularly suitable for standalone embedded solutions.

The MPEG-4 audio standard provides a toolset for audio synthesis and audio processing, i.e. structured audio (SA). SA permits one to describe algorithms through its Structured Audio Orchestra Language (SAOL) programming language. Unlike some other languages of the same type, SAOL has a sample-by-sample execution structure, and this makes particularly important the overhead computation in case of an interpreted decoder implementation. This paper describes the design of an efficient virtual architecture able to exploit the data level parallelism contained in many audio synthesis and processing algorithms and to consistently reduce the implementation overhead through a block-by-block execution.

Despite several efforts aimed at reducing the complexity, block-based motion estimation remains the most computationally demanding stage of video compression algorithms. This is particularly evident when sequences contain large displacements; very large search windows are needed to achieve high-quality coding when such critical conditions occur. This paper presents a new block-motion-estimation algorithm based on the systematic application of the tracing of motion trajectories on a modified genetic search heuristic. The described technique is able to provide motion estimates in very large search windows with optimal coding results. The complexity-reduction factor ranges up to more than two orders of magnitude. The technique can be applied to any macroblock-based video compression standard and to any group of picture structure.