Posts about: Publications

Paper published in the Journal of the Audio Engineering Society, assessment of the quality of distorted audio

7 October 2015

Paper published in the Journal of the Audio Engineering Society by the Good Recording Project team.

For field recordings and user-generated content recorded on phones, tablets, and other mobile devices, poor audio quality arises in part from nonlinear distortions caused by clipping and limiting at pre-amplification stages and by dynamic range control. Based on the Hearing Aid Sound Quality Index (HASQI), a single-ended method to quantify perceived audio quality in the presence of nonlinear distortions has been developed. Validations on music and soundscapes yielded single-ended estimates within ±0.19 of HASQI on a quality range from 0.0 and 1.0. Perceptual tests were carried out to validate the method for music and soundscapes. HASQI has also been shown to predict quality degradations for processes other than nonlinear distortions including additive noise, linear filtering, and spectral changes. By including these other causes of quality degradations, the current model for nonlinear distortion assessment could be expanded.

To go with the publications the authors have also released a program so that if you have some audio that you suspect may be degradaed by amplitude clipping type distortions, this program will be able to detect distorted regions as well as providing a perceptual weighting. Please visit the following link for details on how to acquire the software,



JASA Paper on Perception of Room Modes

30 June 2015

Ever since I was doing my PhD I had a sort of quest to measure the thresholds of modal decay for low frequency room modes. During that time, the guys at Genelec Oy were doing a lot of work on digital equalisation of room/loudspeaker responses. They realised that the cost of digital filters would be greatly reduced if the offending modes could be reduced by a given amount rather than total equalisation. I managed to measure modal thresholds but due to the experimental methodology they were defined in terms of Q-factor rather than decay times for specific frequencies. This was later published in the Journal of the Audio Eng. Society in 2007.

Andrew Goldberg, who was with Genelec at that time was also keen to measure the thresholds and, as is often the case in research, was getting similar results to mine but through different routes and methods. We published some papers at the reproduced sound conference and I clearly remember Andrew whispering to me as we went through the typical rigmarole of swapping speakers between presentations: ‘This is all so close!”.

A few years later, Matthew Stephenson started his PhD with me at the University of Huddersfield and one of the first avenues of research we discussed was the measurement of these thresholds. During Matt’s PhD we approached Andrew and set out to finally measure the thresholds.

The work finally got published in the Journal of the Acoustical Society of America this year (March 2015). The thresholds are defined for critical listening modes, i.e. measured with tones and in isolation of masking, and for more general music listening, using music samples auralised within virtual room responses which include effects of temporal and frequency masking. The measured thresholds were then used to define a map which indicates time-frequency regions where modal decay can be detected by listeners.

The thresholds published allow a more precise target for modal control techniques; passive absorption treatment or loudspeaker equalisation filters that attempt to ameliorate the problems of modes in the room. Depending on budget and user requirements, the different thresholds can be chosen. For rooms where audio programme is listened to in a distracted mode (see Truax’s definition of listening modes), the music thresholds can be used to ensure no odd artefacts from the room modes are perceived. For critical listening spaces such as recording, mastering and post-production studios, the thresholds measured with artificial stimuli are more appropriate but will also require more expensive treatment. Finally, for those who want to completely eradicate low frequency room effects from their listening experience, thresholds that lie at the 2.5% confidence interval of the measured data will ensure that no modal effect will ever disturb it.

A link to the publication:

Bruno Fazenda

Experiments in object based audio

1 December 2014

To cope with the volume of information with which our senses are constantly bombarded, our brains utilise a variety of categorisation strategies to reduce the amount of data it has to process. I’m in the process of setting up some exciting experiments in the acoustics labs at Salford to investigate how categorisation is utilised when our brains process complex acoustic information. I’ll be using methods developed in cognitive psychology to determine how listeners categorise individual sounds in different types of broadcast audio material.

The application of this work is in object based audio, which is the future of broadcast audio. Traditionally, audio content is produced in such a way that the channels of audio information are mapped to a specific loudspeaker layout, such as stereo or 5.1 surround. The limitation of this approach is that the experience of the listener is severely impaired if the reproduction loudspeaker layout does not match the layout for which the audio was produced. Object based audio gets around this by sending each individual audio object (this may be a character’s dialogue, Foley effects, or music) along with information about the object’s position in space and time. Using this information at the receiving end, the audio can be reconstructed in a way that is optimal for the reproduction system; be that headphones, a tablet, or a cinema system. This approach also opens up possibilities for the listener at home to interact with the audio content, which had been explored recently by the BBC, such as choosing which side of the crowd you hear in a football match.

The results of the experiments I’m about to run at Salford will help us to understand what types of objects we need to represent when we store and transmit object based audio, and will lead to experiments exploring the effects different objects types have on the quality of the listener’s experience. This work is part of the S3A project, which is a five year collaborative project involving Salford, Surrey, and Southampton Universities, and BBC R&D that aims to develop immersive 3D audio systems that work in real environments, such as people’s living rooms. The project’s just getting started, so watch this space for more information!

Dr James Woodcock

Microphone wind noise – published in the Journal of the acoustical society of america paper

11 September 2014

Our work into the perception and automated detection of microphone wind noise had been published in the Journal of The Acoustical Society of America. This paper discuss how wind noise is perceived by listeners, and uses this information to form the basis of s wind noise detector / meter for analyzing audio files you can access the Journal here:

Or if you don’t have access, the paper is will also be available here (the next couple of days)

If you want to run the wind noise detection algorithm you can do so using the code here

New paper in AAuA special issue on Auralization and Ambisonics

2 September 2014

Dr Jonathan Hargreaves‘ paper  ‘An energy interpretation of the Kirchhoff-Helmholtz boundary integral equation and its application to Sound Field synthesis’, which was selected as one of the 10 best papers at the  the EAA Joint Symposium on Auralization and Ambisonics in Berlin in April, was published today in Acta Acustica United with Acustica. You can find it via USIR at

Four Salford papers in this month’s edition of JASA!

16 January 2014

It’s a Salford takeover this month in JASA with four papers being published in one issue:


New JASA paper published – simulating acoustic scattering from atmospheric temperature fluctuations

15 January 2014

The Acoustics Research Centre has a strong history in the development and characterisation of SODAR (SOund Detection And Ranging) devices. These devices measure the backscattering of sound pulses transmitted into the lower atmosphere, allowing remote sensing of a variety of data including inversion layers and vertical profiling, wind speed, wind direction, turbulence quantities and stability classes.  Unlike direct measurement techniques (such as mast anemometers) they are quick to deploy and provide continuous data with height; hence they find application in atmospheric research, pollution monitoring and wind-energy surveying.

SODAR and Wind Turbine

In measuring these quantities however SODAR devices face challenges with range and velocity resolution as well as signal to noise ratio problems.  To improve the accuracy of these parameters, new signals and analysis methods need to be evaluated.  This is difficult to achieve by experimentation however, since the ‘true’ atmospheric data required for comparison is not available and must be acquired either from similar instruments or other devices with their own limitations. Existing analytical models of atmospheric scattering are necessarily stochastic in nature, and cannot provide specific data on the response from a given snapshot of atmospheric properties, only information on bulk averages. Thus, there is a requirement for deterministic a SODAR simulator to inform on SODAR performance characteristics over a range of atmospheric conditions.

Model solution process

Model solution process

This new paper by Jonathan Hargreaves, Paul Kendrick and Sabine von Hunerbein, presents a model which combines an analytical incident sound model with a k-space model of the scattered sound close to the inhomogeneous region and a near-to-far-field transform to obtain far-field scattering patterns (depicted above) Results from two test case atmospheres are presented: one with periodic temperature fluctuations with height and one with stochastic temperature fluctuations given by the Kolmogorov spectrum. Good agreement was seen with theoretically predicted far-field scattering and the implications for multi-frequency SODAR design are discussed.

Link to read the paper on USIR:

The work was supported by the UK Engineering and Physical Sciences Research Council [grant number EP/G003734/1 “Advanced Signal Processing Methods Applied to Acoustic Wind Profiling for Use in Wind Farm Assessment”]