You are in: Home » History of the Unit
4. HEARING
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.
4.1 THE DISRUPTION OF AUDITORY PROCESSING BY NOISE (Project 22) (Lutfi, Milroy, Nimmo-Smith, R.Patterson, Weber) (Partial funding from RAE, ISVR, and the CAA.)
It is important to be able to predict the intensity that speech or an auditory warning signal must have for it to be audible in a particular noise environment such as an aircraft cockpit or a hospital ward. Previously, this led R. Patterson and Nimmo-Smith to develop a quantitative model of auditory masking based on the concept of an auditory filter bank; that is, a set of adjacent linear bandpass filters that divides incoming sound into separate frequency channels. The success of the model depends primarily on the accuracy of our representation of the shape of these auditory filters. In this reporting period we determined the shape of the filter as a function of age (284) and stimulus intensity (209), and demonstrated that the shape measurements are not confounded either by the detection of auditory distortion tones (208) or the variability of the masker (379). We also developed a new mathematical description of the filter whose complexity increases as the data require (283) and showed that 1t provides surprisingly precise predictions of the masking levels in the laboratory and on the flight decks of civil aircraft.
The success of this 'Roex' model has led to a set of collaborations with the Royal Aircraft Establishment in Farnborough and the Institute of Sound and Vibration Research in Southampton to predict auditory masking in helicopters -- a particularly difficult type of noise environment. The initial project consisted of three masking experiments including one in a full-scale helicopter simulator, all supervised by APU. It showed that the Roex model could predict listener's threshold to within the accuracy of the noise measurements using population parameter values obtained from the literature. The project report is being prepared currently and includes a microcomputer version of the Roex model to make it more generally available. A direct extension of the project to predict masking in civil helicopters has now been undertaken by the Civil Aviation Authority, again under our supervision. The Navy Applied Psychology Unit at Teddington has asked us to assist in adapting the model to predict masking in naval sonar rooms and the model also provided the basis for a chapter on Voice Communications written for the Navy (278).
A comprehensive review paper that summarises research on the measurement and application of auditory filter-shapes has been written by R. Patterson in collaboration with Dr. Moore of Cambridge University (282).
4.2 THE DESIGN AND EVALUATION OF AUDITORY WARNING SYSTEMS (Project 21) (Edworthy, Milroy, R.Patterson, Shailer) (Partial funding from CAA. RAE, and ISVR.)
The auditory warning systems currently used in transport vehicles and hospitals cause considerable irritation and confusion. There may, for example, be as many as 15 auditory warnings on the flight-deck of a civil aircraft, all of which are too loud, and as many as 50 in the intensive care ward of a hospital, most of which are simple high frequency tones. Previously this led us to determine the appropriate level and number of warnings for a civil aircraft warning system at the behest of the Civil Aviation Authority (CAA). This has generated a series of collaborations in the current reporting period.
A set of guidelines for the design and evaluation of auditory warning systems was written by R. Patterson for the Civil Aviation Authority, which they published as a position paper (275; 277). The principles embodied in the guidelines were illustrated by reviewing the auditory warnings in two current aircraft, the BAC 1-11 (279) and the Boeing 747 (280), and by reviewing an international proposal for auditory warnings in future aircraft (275). The guidelines have since been used to develop a CAA standard for auditory warnings on airlines flying into Britain.
We drew the MRC's attention to the commercial potential of our auditory warnings research and together we filed four patent applications for auditory warning systems through the British Technology Group (BTG) (UK 8222029, UK 83202U4, Europe 83304350.8, and USA 515501). A license to produce warning systems under the patent was taken out by Racal Acoustics and it has generated about £20,000 to date.
We developed a standard for auditory warning systems in hospitals in conjunction with a group of consultant anaesthetists operating under the auspices of the British Standards Institute. This work precipitated the formation of a working group of the International Standards Organisation with whom we have prepared a draft international standard for the Rationalisation of Auditory Warnings in Hospitals. The APU guidelines form the basis of all documents and they are about half way through the standardisation process.
The Directorate of Helicopter Projects (MOD) have asked us through RAE Farnborough to prepare a specification for a standardised set of auditory warning signals in military helicopters based on our masking studies and our guide!ines work.
The warning sounds specified 'in the draft standards of the CAA, the BSI/IS0, and DHP are unlike any previous warnings and it is difficult to describe them in words. As a result, all three groups have asked APU to prepare demonstration warning sets to illustrate and promote the relevant standards. These warning sets will also be used for field trials in hospitals and aircraft, and with the appropriate modifications, will probably become the first of the new generation of computer based auditory warning systems. The OHP warnings are essentially completed, the CAA warnings are in preparation, and funding for the hospital warnings appears imminent.
The collaboration associated with Project Numbers 22 and 21 have been particularly important for the Auditory Group at APU: Firstly, they enabled us to test the Roex model via laboratory and field experiments that we were not able to perform technically and for which we did not have the staff. Secondly, they enabled us to extend the influence of our research and theories in a controlled way, again without taking on extra staff.
Finally, they provided the resources to expand the auditory laboratory at APU over this period (roughly £22,000).
4.3 IMPROVED AUDIOMETRY ASSESSMENT (Project 3) (Terminated 31.3.84) (Milroy, Nimmo-Smith, R.Patterson, Weber)
It is generally assumed that the standard hearing test, the audiogram, should be accompanied, if not surplanted, by a test that reflects frequency resolution rather than just absolute sensitivity. As a result, we have developed a clinical version of the auditory-filter measure and carried out a population study to establish norms for the Roex filter shape for middle-aged and older people (284). The study was extended to show that the filter-shape measure could predict speech intelligibility results better than the audiogram (283).
Subsequently, we have collaborated with Cambridge University to establish a more powerful test that can be used to measure asymmetry in patients' filter shapes (116), and with the MRC Institute of Hearing Research in Nottingham to demonstrate that some patients have abnormally asymmetric filters (335). In particular, we found patients who suffer from the downward, rather than the upward, spread of masking and who would actually be hindered rather than helped by standard hearing aids.
4.4 PSYCHOLOGY OF MUSIC (Project 50) (Edworthy, R.Patterson)
The ability of musicians and non-musicians to discriminate between major and minor chords is currently being studied by Edworthy and R. Patterson. Reasons for this study are twofold. First, the major/minor dichotomy warrants investigation because of its status in Western music; secondly, it 1s being used to study the interaction between key and interval information at a cognitive level, for which no adequate theory currently exists.
When we listen to a melody, which repeats at a different pitch, it may either preserve the key, in which case the intervals must change, or preserve the intervals, in which case the key must change. Key and interval information are mutually dependent as a key cannot be established without intervals, and intervals appear to be more precise when a key can be established (96; 97). Examples of both sorts of invariance commonly occur in music.
There are specific instances of pitch changes for which a preservation of key results in a mode change from major to minor, or vice versa. These changes do not appear 'incorrect' because the key has been preserved. We have crystallised this phenomenon into a series of two-chord Interference studies, where a sense of key can be induced by fixing the pitch of the second chord. Initial data show that when a mode change occurs, but the key is preserved, judgments of mode are more difficult to make than when there is no such conflict of tonal and modal information.
Further experimentation will develop this paradigm to include longer chord sequences, and melodic sequences.
4.5 CENTRAL AUDITORY PROCESSING (Spiral Processing of Sound) (Project 51) (Limbert, Lutfl, Milroy, Nimmo-Smith, R. Patterson, Peters)
During this reporting period R. Patterson discovered a 'spiral' method of extracting pitch from streams of auditory nerve pulses, which appears to be more sensitive and more economical than current methods. It has important implications for both hearing theory and automatic speech recognition and so a new project was initiated to pursue the discovery.
About 10 years ago three different models of pitch perception were published all of which suggested that the mechanical spectral analysis performed by the cochlea might be sufficient to account for the majority of pitch-perception phenomena. Although these three models still dominate hearing theory, it has become apparent that the timing information observed in the phase-locked fibers of the auditory nerve also plays a role in determining pitch. Indeed, it now seems that this temporal information may be the more important factor in the case of musical pitch and speech pitch, and that the primary purpose of the cochlea is to alleviate masking; that is, to prevent the disruption of high-frequency signals by low-freqency noise and vice versa. As a result, attention is now focused on auditory neural processing in the hope that it will provide the basis for a better understanding of speech recognition processes and music perception.
There is a stark contrast in hearing research between our understanding of the mechanical analysis of sound performed by the cochlea and the neural analysis of sound performed by the brain. Whereas there are competent simulations of basilar membrane motion to explain how the cochlea performs its spectral analysis, there are only primitive histogram models to suggest how the brain might extract pitch from the regularly spaced streams of nerve pulses observed in phase-locked auditory fibers when the ear is stimulated by a periodic sound.
Several years ago R. Patterson discovered a spiral mechanism that can convert the timing information in a regularly spaced stream of pulses into a spatial pattern of spokes radiating from the centre of the spiral. This Spiral Processor appears to have the right properties for extracting pitch from neural impulse streams quickly and passively without the need of harmonic templates or sieves. Briefly, the time line along which the nerve pulses flow is wrapped into a logarithmic spiral with base 2, and the nerve pulses flow from the centre of the spiral outwards. For a periodic wave, once per cycle, the nerve pulses coalesce onto a specific pattern of spokes and the orientation of the pattern determines the pitch.
The Spiral Processor appeared to represent a major breakthrough for automatic speech recognition machines, the commercial market for which is estimated to be in the hundreds of millions of pounds in the next decade. To demonstrate the advantages of the Spiral Processor a thirty-filter simulation of the cochlea was programmed on our sound computer and used to convert incoming sounds into the types of pulse streams observed in primary auditory nerves. A thirty channel Spiral Processor was then programmed on the same computer to extract pitch from the ensemble of pulse streams and a dynamic display of the pulse streams flowing along thirty concentric spirals was assembled.
For commercial reasons the project was kept strictly confidential until it was presented to the Industrial Liaison Group of the MRC in June of 1984. A patent application for the spiral sound processor was then developed and filed on January 2nd, 1985 to establish our commercial priority. On January 3rd, 1985 a paper on the spiral processor was presented at a meeting of the Experimental Psychology Society to establish our scientific priority in this area.
Research on pitch and periodicity perception had already begun in the current reporting period before the Spiral Processor was discovered. A pair of studies were performed to show that low pitches like those associated with the vowels of speech are much easier to perceive in short sounds when the fundamental is accompanied by higher harmonics, a finding that supports a temporal theory of pitch perception (285). A subsequent series of experiments on the detection of a repeating tone burst presented in a repeating noise indicated that the high-resolution version of incoming sound produced by the cochlea is reduced to a much lower resolution form within 100 ms of its capture (281). Finally, a series of experiments was performed to investigate the perception of repeating noise and determine whether spectral theories were correct in their assumption that the perception of repetition is simply based on the recurrence of a local maximum in the short-term power spectrum of the repeating noise. The data make it clear that the spectral theory is decidedly incorrect but they did not suggest an alternative temporal mechanism (195).
Other sections in the 1981-1984 report
2. COGNITIVE PSYCHOLOGY
3. COGNITIVE ERGONOMICS/APPLIED COGNITIVE PSYCHOLOGY
4. HEARING
5. MOTOR SKILL AND ACTION
6. VISUAL PERCEPTION
7. PSYCH0PHYSIOL0GY SECTION