AES Rome 2013
Paper Session P4
P4 - Audio Signal Processing—Part 2
Saturday, May 4, 14:30 — 16:30 (Sala Foscolo)
Leonardo Gabrielli, Universitá Politecnica delle Marche - Ancona, Italy
P4-1 User-Driven Quality Enhancement for Audio Signal—Danilo Comminiello, Sapienza University of Rome - Rome, Italy; Simone Scardapane, Sapienza University of Rome - Rome, Italy; Michele Scarpiniti, Sapienza University of Rome - Rome, Italy; Aurelio Uncini, Sapienza University of Rome - Rome, Italy
Classical methods for audio and speech enhancement are often based on error-driven optimization strategies, such as the mean-square error minimization. However, these approaches do not always satisfy the quality requirements demanded by users of the system. In order to meet subjective specifications, we put forward the idea of a user-driven approach to audio enhancement through the inclusion in the optimization stage of an interactive evolutionary algorithm (IEA). In this way, performance of the system can be adapted to any user in a principled and systematic way, thus reflecting the desired subjective quality. Experiments in the context of echo cancellation support the proposed methodology, showing significant statistical advantage of the proposed framework with respect to classical approaches.
Convention Paper 8823 (Purchase now)
P4-2 Partial Spectral Flatness Measure for Tonality Estimation in a Filter Bank-Based Psychoacoustic Model for Perceptual Audio Coding—Armin Taghipour, International Audio Laboratories Erlangen - Erlangen, Germany; Maneesh Chandra Jaikumar, International Audio Laboratories Erlangen - Erlangen, Germany; Hochschule Rosenheim, University of Applied Science - Rosenheim, Germany; Bernd Edler, International Audio Laboratories Erlangen - Erlangen, Germany; Holger Stahl, Hochschule Rosenheim, University of Applied Science - Rosenheim, Germany
Perceptual audio codecs use psychoacoustic models for irrelevancy reduction by exploiting masking effects in the human auditory system. In masking, the tonality of the masker plays an important role and therefore should be evaluated in the psychoacoustic model. In this study a partial Spectral Flatness Measure (SFM) is applied to a filter bank-based psychoacoustic model to estimate tonality. The Infinite Impulse Response (IIR) band-pass filters are designed to take into account the spreading in simultaneous masking. Tonality estimation is adapted to temporal and spectral resolution of the auditory system. Employing subjective audio coding preference tests, the Partial SFM is compared with prediction-based tonality estimation.
Convention Paper 8824 (Purchase now)
P4-3 A New Approach to Model-Based Development for Audio Signal Processing—Carsten Tradowsky, Karlsruhe Institute of Technology (KIT) - Karlsruhe, Germany; CTmusic - Karlsruhe, Germany; Peter Figuli, Karlsruhe Institute of Technology (KIT) - Karlsruhe, Germany; Erik Seidenspinner, Karlsruhe Institute of Technology (KIT) - Karlsruhe, Germany; Felix Held, Karlsruhe Institute of Technology (KIT) - Karlsruhe, Germany; Jürgen Becker, Karlsruhe Institute of Technology (KIT) - Karlsruhe, Germany
Today, digital audio systems are restricted in their functionality. For example, a digital audio player still has a resolution of 16-bit and a sample rate of 44.1 kHz. This relatively low quality does not exhaust the possibilities given by modern hardware for music production. In most cases, the functionality is described in software. This abstraction is very common these days, as only few engineers understand the potential of their target hardware. The design-time increases significantly to develop efficiently for the target hardware. Because of the use of common compiler tool chains the software is statically mapped onto the hardware. This restricts the number of channels per processing core to a minimum when targeting high quality audio. One possibility to close the productivity gap, described above, is to use a high-level model-based development approach. The audio signal processing flow is described in a more abstract high level using the model-based development approach. This model is then platform-independently compiled including automatically generated simulation and verification input. Platform-dependent code can be automatically generated out of the verified model. This enables the evaluation of different target architectures and their trade-offs using the same model description. This paper presents a concept to use a model-based approach to describe audio signal processing algorithms. This concept is used to compile C- and HDL-code out of the same model description to evaluate different target platforms. The goal of this paper is to compare trade-offs for audio signal processing algorithms using a multicore Digital Signal Processor (DSP) target platform. Measurements using data parallelism inside the generated code show a significant speedup on the multicore DSP platform. A conclusion will be made regarding the usability of the proposed model-based tool flow as well as the applicability on the multicore DSP platform.
Convention Paper 8825 (Purchase now)
P4-4 Accordion Music and its Automatic Transcription to MIDI Format—Tomasz Maciejewski, Poznan University of Technology - Poznan, Poland; Ewa Lukasik, Poznan University of Technology - Poznan, Poland
The paper is devoted to the problems related to the automatic transcription of the accordion sound. The accordion is a musical instrument from the free-reed aerophone family that is able to produce polyphonic, multi-chord music. First the playing modes are briefly characterized and problems related to the polyphonic nature of the sound is discussed. Then the results of the analysis and MIDI transcription of the right-side monophonic and polyphonic melodies are presented. Finally, an attempt to transcribe music generated by both sides of the instrument recorded in two channels is presented giving the foundation to further research.
Convention Paper 8827 (Purchase now)