You are in: Home » History of the Unit
AUDITION
General Notes. This material has been scanned from the original typescript. While we have done our best to remove errors, some may well remain. You can access other parts of this particular Progress Report either from the menu at the bottom of this entry, by using your browser's back function, by navigating back to the Unit History Timeline, or by accessing the relevant section of the electronic archive. Reference for this report are indexed by number and these can be found in a dedicated section also accessible from the menu at the bottom of this entry.
During the previous and current reporting periods, hearing research at APU was divided into three roughly equal areas — auditory warnings, auditory filtering and auditory perception. During the current reporting period the work on filtering and perception merged, and a project to develop a complete model of monaural auditory processing was begun. As a result, auditory research will be described under two project titles: 60. Analysis and Modelling of Auditory Perception, and 61. Auditory Warnings.
Project 60 - Analysis and Modelling of Auditory Perception
The auditory model under development at the APU has four stages: The first two stages perform spectral analysis via an auditory filterbank and neural transduction via a bank of "hair cells". Together, the first two stages form a cochlea simulation. The third stage performs phase alignment and produces a high resolution, stabilised image of the sound quality, or timbre. The fourth stage extracts the pitch and forms a global stabilised image whose motion corresponds to the pitch contours that we hear in music and speech. Together, the latter two stages simulate the auditory neural processing that converts the data flowing from the cochlea into a simulation of human auditory sensation. During the current reporting period Patterson has developed a complete computational version of this auditory model that can record and analyse the sound into pitch and timbre images without intervention [43]. During the next reporting period the intention is a) to improve the scientific basis of the model through experiments on pitch and timbre perception, b) to develop an auditory workstation and an auditory PC to promote the use of the model, and c) to develop collaborations to evaluate the model as a preprocessor for speech recognition.
60.1 Auditory Filterbank (Cosgrove, Edworthy, Holdsworth, Milroy, Nimmo-Smith, R Patterson)
In collaboration with the Royal Aircraft Establishment, Farnborough and the Institute of Sound and Vibration Research, Southampton, Patterson et al demonstrated that the Roex model of auditory masking could be used to predict signal threshold in complex noise environments like helicopters and trains [59, 55]. As a result, the collaboration was extended with the mandate to develop a time-domain, dynamic version of the filter, and an auditory filterbank that could serve as a preprocessor for speech sounds. This led Patterson and Holdsworth to develop a "gammatone auditory filterbank", which has the following advantages: The filter function was developed by auditory physiologists and its temporal characteristics have been shown to mimic those observed in cat data. When tuned to human parameters, the amplitude characteristic of the gammatone filter provides an extremely good approximation to that of the Roex filter which ensures that it will predict human masking data as well as the Roex filter. Finally, Holdsworth discovered a recursive implementation of the gammatone filter that makes the computation fast enough to be used as a speech preprocessor.
Following the development of a convenient interface, the filterbank has been distributed to about 20 sites in Britain and North America [63].
60.2 Neural Transduction (R Patterson)
At the start of this reporting period, Patterson programmed a simple static neural transduction mechanism that was sufficient to convert the output of the original auditory filterbank into the pulse ribbons used in our pitch and timbre research [47]. With the advent of the gammatone filterbank and more professional programming skills, a dynamic neural transducer was imported from Loughborough, creating our first truly viable cochlea simulation. These neural transduction mechanisms were used in conjunction with a lengthy series of experiments on monaural phase perception to produce a model of human phase perception that relates timbre changes to global parameters of auditory nerve firing [44, 56]. The Loughborough transduction mechanism includes compression and adaptation like those in the auditory system but it does not include the lateral suppression observed in the auditory system. As a result, the output of the mechanism tends to blur auditory features rather than sharpening them. This prompted Holdsworth to consider the form of the signal processing applied during neural transduction and to generate a mechanism which at one and the same time is simpler and includes suppression. This "suppressive transducer" sharpens auditory features like the formants of speech. It is also particularly efficient and would appear to have considerable potential as a component in a speech preprocessor. A paper describing the "suppressive transducer" has been written and used as a basis for a patent application.
60.3 The Stabilised Timbre Image (R Patterson)
When a sound is periodic the neural firing pattern at the output of the cochlea contains a repeating pattern. The details of the pattern describe the timbre of the sound and the best description of the timbre is provided by combining the information from successive cycles. The original version of the auditory model [47] which suggested wrapping the output of the cochlea around a cylinder to concentrate the timbre information, proved to be computationally intractable. Subsequently, Patterson developed a triggered, quantised method of temporal integration that preserves the fine-grain temporal information during the longer term integration process. The mechanism is passive and produces a stabilised auditory image for periodic sounds the way the auditory system does. A computational version of this third stage now exists and a refined version of the first three stages of the model will form the initial auditory preprocessor that will be provided to the collaborators in the Esprit BRA project. A paper describing "quantised temporal integration" has been written and used as a basis for a patent application.
The work on the stabilised auditory image has attracted the interest of the Royal Aircraft Establishment, Farnborough, and we have negotiated a substantial research grant to develop the mechanism for speech stimuli.
60.4 The Stabilised Pitch Image (Nimmo-Smith, R Patterson)
In the model, the pitch information is extracted from the pulse ribbon flowing from the cochlea by wrapping the time line of the pulse ribbon into a logarithmic spiral, base 2. A periodic sound causes a subset of the pulses to line up on a specific spoke of the spiral and the orientation of the spoke determines the pitch of the sound [43]. Patterson and Nimmo-Smith investigated the properties of the spiral processor as a formal detection process and developed a staged argument to illustrate how this new, highly non-linear, detector is related to more traditional Fourier detectors and why it is more efficient for processing sounds like music and speech [51]. The MRC has applied for and received patent protection for the spiral processor [61].
Project 61 - Design and Evaluation of Auditory Warning Systems
Auditory warnings are used throughout industry because hearing is a 360° warning sense; no matter where the operator is looking, the warning sound will be detected. The need for auditory warnings was tragically exemplified by the crash of a British Airways passenger helicopter which did not have an auditory warning. While the pilots were looking for landfall in fog, the helicopter slowly descended into the sea. There was a flashing yellow warning light but they did not see it. An auditory low-height warning would almost certainly have averted this accident. In other industries, the problem is not a lack of auditory warnings but a plethora of excessively loud and strident warning sounds.
In the last reporting period, at the request of the Civil Aviation Authority, Patterson developed a set of guidelines for the production of auditory warning systems in aircraft and showed how the guidelines could be used to review four existing and proposed auditory warning systems. Subsequently specifications for three warning systems were prepared for use in fixed-wing civil aircraft, in intensive care wards of hospitals and in military helicopters.
During the current reporting period, five auditory warning systems have been developed, two were designed to support British and International Standards; the remaining three were designed for use in specific applications [46, 48]. The warning sounds have the following general characteristics: A burst of three to eight sound pulses is used to construct a brief warning of 1-2 seconds duration. Each warning has a distinctive rhythm and melody, to prevent its being confused with other members of the warning set, and it is played with varying rates and keys to indicate the urgency of the situation. The spectrum of the sound pulse ensures that the warning sound is audible in the environment for which it is designed and that it has a distinctive sound quality. The following five paragraphs provide a brief description of the purpose and production of each warning set.
61.1 British and International Standards for Aircraft Warnings (R. Patterson)
At the request of the Civil Aviation Authority, a set of nine auditory warnings was developed for use in civil airliners. The purpose of this set of demonstration warnings was to illustrate the kind of warning sounds that would satisfy a Standard being developed by the Civil Aviation Authority for use in Britain and Europe. One of the primary purposes of this set of warnings was to illustrate that the design principles would not constrain manufacturers unduly and that the warnings would be highly distinctive. The warning set was the topic of an invited lecture presented to the Royal Aeronautical Society.
61.2 British and International Standards for Hospital Warnings (Edworthy, R Patterson)
At the request of the British Standards Institute and with the support of the Department of Trade and Industry, a set of warning sounds was developed for use in the operating theatres and intensive care wards of hospitals [53, 62]. The purpose of the warning set was to illustrate to British and International Standards organizations the kind of civilised, distinctive warning sounds that could be specified in a Standards document and used to replace the cacophony of buzzers and bells used currently. The standardisation process has now proceeded to the level of a Draft International Standard and should eventually become a Standard. The Standard specifies, and the demonstration warnings illustrate, two forms of hospital warning system. In one case there are only three sounds each of which indicates a whole category of problems and which are differentiated by their urgency. In the second form the three category sounds are supplemented by six specific warnings all of which indicate urgent conditions occurring in the topmost category. The design represents a compromise that enables each authority or hospital ward to tailor the system to their own needs by adding a small number of the highly urgent warning sounds to the general set of three category warnings.
61.3 Military Helicopter Warnings (Edworthy, Milroy, R Patterson)
This was a large project that extended over the entire reporting period and resulted in the production of three warning sets for three different helicopters (Sea King, Lynx, and Puma) [59]. The first set was developed for the multi-role Sea King helicopter and consisted of ten warning sounds. Like the civil aircraft set, it was designed to illustrate the wide range of distinctive sounds that could be constructed from within the guidelines. A learning experiment was performed with helicopter pilots and it showed that the new sounds were much more resistant to confusion than those being used currently in civil airliners [58]. The second set of warnings was produced for the Lynx helicopter and used to check the guidelines for setting the sound levels of the warning sounds. Detection levels were measured in a Lynx helicopter shell and it showed thatPatterson's roex model of auditory masking was as accurate as the noise measurements that could be made beside the pilot's ear. In the course of these two studies, the helicopter pilots requested more urgent-sounding warnings to reflect the fast pace of life in the low-flying helicopter. Accordingly, a third and final set of warning sounds was developed for the Puma helicopter. Following tests to ensure the detectability and discriminability of these new warnings, they have been installed in test helicopters with two operational squadrons.
61.4 Civil Helicopter Warnings (R Patterson)
Following the crash of a British Airways helicopter, the Civil Aviation Authority asked the APU to produce a demonstration set of warnings for helicopters ferrying staff and supplies to North Sea oil rigs. The warnings were demonstrated to the operators of North Sea Helicopters who chose a modified subset of the sounds to meet a Civil Aviation Authority requirement for low-height warnings. APU and Racal Acoustics implemented the warning sounds in an airworthy device, and they are currently installed in more than 150 helicopters. A patent for warning sound production was submitted in the previous reporting period and has now been granted by British, European and American patenting authorities [60].
61.5 British Rail Trackside Warnings (Cosgrove, Milroy, R Patterson)
At the request of British Rail Research (Derby), a set of warning sounds for use by trackside maintenance crews to warn of approaching trains has been designed and tested. The correspondence between trackside warning function and the existing sound has been preserved by constructing all of the new warning sounds from components in the existing sound [57]. All four warning sounds had to be audible in the presence of no less than 46 different noise environments specified by British Rail Research [55]. A learning experiment was performed to show that the warning sounds are highly resistant to confusion [57]. The sounds have been delivered to British Rail in the form of an annotated demonstration tape; trackside testing will begin sometime later this year.
REFERENCES Al - Authored Books
A2 - Edited Books
41. Moore, B.C.J, and PATTERSON, R.D. (Eds.) (1986) Auditory Frequency Selectivity, New York: Plenum Publishing Corporation, NATO ASI Series A: Life Sciences Vol. 119.
B - Refereed Journal Articles
42. EDWORTHY, J. (1985) Interval and contour in melody processing. Music Perception, 2, 375-388.
43. PATTERSON, R.D. (1986) Spiral detection of periodicity and the spiral form of musical scales. Psychology of Music, 14, 44-61.
44. PATTERSON, R.D. (1987) A pulse ribbon model of monaural phase perception. Journal of the Acoustical Society of America, 82, 1560-1586.
C - Invited Chapters and Commentaries
45. EDWORTHY, J. (1985) Melodic contour and musical structure. In P.Howell, I. Cross and R. West (Eds.), Musical Structure and Cognition. London: Academic Press Ltd., pp.169-188.
46. PATTERSON, R.D. (1985) Auditory warning systems for high-workload environments. In I.D. Brown, R. Goldsmith, K. Coombes and M.A. Sinclair (Eds.), Ergonomics International 85. London: Taylor and Francis, pp.163-166.
47. PATTERSON, R.D. (1987) A pulse ribbon model of peripheral auditory processing. In W.A. Yost and C.S. Watson (Eds.), Auditory Processing of Complex Sounds. Hillsdale, N.J.: Lawrence Erlbaum Associates, pp.167-179.
48. PATTERSON, R.D. (1989) Guidelines for the design of auditory warning sounds. In Proceedings of Acoustics '89 Spring Conference, Vol.11 (5). Edinburgh: Institute of Acoustics, pp. 17-24.
49. PATTERSON, R.D., and CUTLER, A. (1989) Auditory preprocessing and recognition of speech. In A.D. Baddeley and N.O. Bernsen (Eds.), Research Directions in Cognitive Science: A European Perspective, Vol. 1: Cognitive Psychology. London: Lawrence Erlbaum Associates, pp.23-60.
50. PATTERSON, R.D. and Moore, B.C.J. (1986) Auditory filters and excitation patterns as representations of frequency resolution. In B.C.J. Moore (Ed.), Frequency Selectivity in Hearing. London: Academic Press Ltd., pp.123-177.
51. PATTERSON, R.D. and NIMMO-SMITH, I. (1986) Thinning periodicity detectors for modulated pulse streams. In B.C.J. Moore and R.D. Patterson (Eds.), Auditory Frequency Selectivity (NATO ASI Series A: Life Sciences, Vol. 19). New York: Plenum Publishing Corporation, pp.299-307.
D - Conference Proceedings
52. COSGROVE, P., Wilson, J.P. and PATTERSON, R.D. (1989) Formant transition detection in isolated vowels with transitions in initial and final position. In Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing '89, Vol.1 (Section Si). Glasgow, Scotland: Bell and Brian Ltd., pp.278-281.
53. EDWORTHY, J. and PATTERSON, R.D. (1985) Ergonomic factors in auditory warnings. In I.D. Brown, R. Goldsmith, K. Coombes and M.A. Sinclair (Eds.), Ergonomics International 85. London: Taylor and Francis, pp.232-234.
54. LOGIE, R.H. and EDWORTHY, J. (1986) Shared mechanisms in the processing of verbal and musical material. In D.G. Russell, D.F. Marks and J.T.E. Richardson (Eds.), Imagery 2. Proceedings of the 2nd International Imagery Conference, Swansea, April 1985. Dunedin, New Zealand: Human Performance Associates, pp.33-37.
55. Lower, M.C., PATTERSON, R.D., COSGROVE, P. and MILROY, R. (1989) Sound levels for the British Rail Inductive Loop Warning System. In Proceedings of Acoustics '89 Spring Conference, Vol.11 (5). Edinburgh: Institute of Acoustics, pp.43-50.
56. PATTERSON, R.D. (1988) Timbre cues in monaural phase perception: Distinguishing within-channel cues and between-channel cues. In H. Duifhuis, J.W. Horst and H.P. Wit (Eds.), Basic Issues in Hearing. Proceedings of the 8th International Symposium on Hearing. London: Academic Press Ltd., pp.351-358.
57. PATTERSON, R.D., COSGROVE, P., MILROY, R. and Lower, M.C. (1989) Auditory warnings for the British Rail Inductive Loop Warning System. In Proceedings of the Acoustics '89 Spring Conference, Vol.11 (5). Edinburgh: Institute of Acoustics, pp.43-50.
58. Shailer, M.J. and PATTERSON, R.D. (1985) Pulse generation for auditory warning systems. In I.D. Brown, R. Goldsmith, K. Coombes and M.A. Sinclair (Eds.), Ergonomics International 85. London: Taylor and Francis, pp.229-231.
E - Technical Reports, Theses and Tests
59. Lower, M.C., PATTERSON, R.D., EDWORTHY, J., Shailer, M.J., MILROY, R. and Wheeler, P.D. (1986) The design and production of auditory warnings for helicopters. Institute of Sound and Vibration Research Report AC527A.
60. PATTERSON, R.D. (1985) Apparatus and Methods for Generating Auditory Indicators. London: UK Patent Office, Patent No. GB 2124417 B.
61. PATTERSON, R.D. (1988) Analysis of Non-Sinusoidal Waveforms. London: UK Patent Office, Patent No. GB 2169719 B (1988).
62. PATTERSON, R.D., EDWORTHY, J., Shailer, M.J., Lower, M.C. and Wheeler, P.D. (1986) Alarm sounds for medical equipment in intensive care areas and operating theatres. Institute of Sound and Vibration Research Report No. AC598.
63. PATTERSON, R., NIMMO-SMITH, I., HOLDSWORTH, J. and Rice, P. (1988) Spiral VOS Final Report Part A: The Auditory Filter Bank. CED Contract Report.
Other sections in the 1985-1989 report
3. LANGUAGE, SPEECH, READING AND WRITING