AES Berlin 2014
Paper Session P3
P3 - Signal Processing—Part 1
Sunday, April 27, 09:00 — 11:30 (Room Paris)
Ville Pulkki, Aalto University - Espoo, Finland; Technical University of Denmark - Denmark
P3-1 Memory Requirements Reduction Technique for Delay Storage in Real Time Acoustic Cameras—Héctor A. Sánchez-Hevia, University of Alcalá - Alcalá de Henares, Madrid, Spain; Inma Mohino-Herranz, Universidad de Alcalá - Alcalá de Henares, Madrid, Spain; Roberto Gil-Pita, University of Alcalá - Alcalá de Henares, Madrid, Spain; Manuel Rosa-Zurera, University of Alcalá - Alcalá de Henares, Madrid, Spain
Acoustic cameras are devices capable of displaying a visual representation of sound waves. Typically these devices relay on delay-based techniques, such as Delay and Sum Beamforming, being the calculation of the proper delay values a key component of the system. For real-time systems with a large amount of microphones it is not practical to perform such calculation being common to go for an offline strategy in which the pre-calculated values are stored in memory, allowing a faster dispatch of the data while increasing memory requirements. In this paper we present a technique for delay storage optimization based on various symmetries found within the pre-calculated values that allow a reduction up to almost 16 times over the initial memory requirements.
Convention Paper 9031 (Purchase now)
P3-2 Introducing Waveform Distribution Moments for Audio Content Analysis—Henrik von Coler, SIM (Staatliches Institut für Musikforschung) - Berlin, Germany; Technical University of Berlin - Berlin, Germany
This paper introduces waveform distribution moments as features for audio content analysis. Moments and central moments of distributions are directly calculated from the squared waveform, in order to extract information on the energy development of a signal. The feature trajectories thus obtained promise to be applicable in transient detection, onset detection, and related tasks and are more sensitive to rapid changes than root mean square based methods, as a qualitative analysis reveals. An evaluation of the proposed features is presented in a feature ranking experiment related to transient detection and in an onset detection experiment. In both applications the waveform distribution moments show promising results in comparison to other signal descriptors.
Convention Paper 9032 (Purchase now)
P3-3 Efficient Low Frequency Echo Cancellation Using Sparse Adaptive FIR Filters—Alexis Favrot, Illusonic GmbH - Lausanne, Switzerland; Christof Faller, Illusonic GmbH - Uster, Switzerland
It is shown how finite impulse response (FIR) filtering and filter adaptation can be implemented with reduced computational complexity when applied to signals containing only low frequencies. A sparse adaptive filter (with only every Mth coefficient being non-zero) with reduced adaptation rate achieves a similar result as a conventional adaptive filter but with lower computational complexity. An echo control scheme based on a sparse adaptive filter is described. Low frequency echoes are cancelled followed by echo suppression over all frequencies.
Convention Paper 9033 (Purchase now)
P3-4 Computationally-Efficient Speech Enhancement Algorithm for Binaural Hearing Aids—David Ayllón, University of Alcalá - Alcalá de Henares, Spain; Roberto Gil-Pita, University of Alcalá - Alcalá de Henares, Madrid, Spain; Manuel Rosa-Zurera, University of Alcalá - Alcalá de Henares, Madrid, Spain
The improvement of speech intelligibility in hearing aids is a complex and unsolved problem. The recent development of binaural hearing aids allows the design of speech enhancement algorithms to take advantages of the benefits of binaural hearing. In this paper a novel source separation algorithm for binaural speech enhancement based on supervised machine learning and time-frequency masking is presented. The proposed algorithm requires less than 10% of the available instructions for signal processing in a state-of-the-art hearing aid and obtains good separation performance in terms of WDO for low SNR.
Convention Paper 9035 (Purchase now)
P3-5 Two-Stage Impulsive Noise Detection Using Inter-frame Correlation and Hidden Markov Model for Audio Restoration—Kwang Myung Jeon, Gwangju Institute of Science and Technology (GIST) - Gwangju, Korea; Dong Yun Lee, Gwangju Institute of Science and Technology (GIST) - Gwangju, Korea; Nam In Park, Gwangju Institute of Science and Technology (GIST) - Gwangju, Korea; Myung Kyu Choi, Samsung Electronics - Gyeonggi-do, Korea; Hong Kook Kim, Gwangju Institute of Science and Tech (GIST) - Gwangju, Korea
In this paper a two-stage impulsive noise detection method is proposed to improve the quality of audio signals distorted by impulsive noise. In order to reduce false alarms and missing detection errors, the proposed method first tries to detect whether a frame includes onsets on the basis of inter-frame correlation. Next, hidden Markov model-based maximum likelihood classification is carried out to decide if the onset has occurred from impulsive noise or not. It is shown from performance evaluation that the proposed method achieves higher detection accuracy than with conventional residual domain-based methods under various impulsive noise distributions.
Convention Paper 9036 (Purchase now)