Professor John F Culling - BSc DPhil Sussex
Listeners are highly proficient at detecting and identifying sounds, especially speech sounds, in background noise. This ability is remarkable, because the waveform of the attended voice may be quite swamped by those of competing voices at the two ears. It also has important practical ramifications, since hearing impaired listeners often find a single voice intelligible when amplified, but find any interfering sound intolerable. In very noise environments normally-hearing listeners will also struggle, especially in reverberant rooms. By investigating the perceptual mechanisms which underlie these effects, I hope to uncover principles which could guide the design of hearing-aids, cochlear implants and, indeed, rooms so that they facilitate rather than impede communication in noise.
I teach Normal and Impaired Hearing at level 3 (PS3308). My lectures examine the auditory system from a mainly psychoacoustic perspective, explaining how the functions of the cochlea and brain can be elucidated purely through listening experiments. The physiology of the ear is mainly presented towards the end of the lecture series as an introduction to the effects hearing impairment and nature of the hearing facilitated by cochlear implantation. I also supervise level 2 practicals in perception, focussing on the parameters of the human voice, their sexual differentiation and the potential influence of these parameters on vocal attractiveness.
Selected publications (2014 onwards)
Research topics and related papers
My research focuses on the cocktail-party problem, which concerns how listeners are able to cope with high levels of interfering noise when listening to speech. Typically, such interfering noise may consist of many other competing voices, as at a cocktail party or a busy restaurant. Humans (and other animals) remain far more proficient at this task than any automatic system. They are known to use many different mechanisms, all of which fall within my research interests, but most of my research concerns binaural hearing.
Much of my early work was concerned with differences in F0 (DF0s). I found that the effect of DF0s operates principally in the first formant region (Culling and Darwin, 1993), and that DF0s produce multiple perceptual cues, including amplitude modulations (Culling and Darwin, 1994). Differences in modulation of F0 were only useful when that modulation introduced a difference in F0 (Darwin and Culling, 1990).
Through the possession of two ears listeners are able to exploit differences in sound source location. The effect involves interaural level differences (ILDs) and interaural time differences (ITDs), the cues used for left/right sound localisation (Culling et al., 2004). However, distinct perceived locations for target and interfering sound are not necessary or sufficient for listeners to perceptually separate competing sounds. Culling and Summerfield (1995) found that listeners were unable to perceptually segregate a whispered vowel, represented by two noise bands with a common ITD, from two concurrent (but spectrally distinct) noise bands with a different ITD. Edmonds and Culling (2005a) showed that speech need not have a consistent ITD in different frequency bands for listeners to perceptually segregate it from competing speech or noise, and that ITDs and ILDs need not even agree regarding source direction to get the full benefit of both cues (Edmonds and Culling 2005b).
Rather than relying on sound-source localisation, the binaural system seems to use a separate process, known as binaural unmasking. It has long been known that a signal can be detected or identified at a lower signal-to-noise ratio (SNR) if its interaural timing or phase differs from that of a masker. A similar effect is observed for the intelligibility of speech signals. One framework used to explain binaural unmasking is interaural cancellation, which was originally developed by Nat Durlach but which has lately been adapted for use with broadband signal such as speech (Culling and Summerfield, 1995; Culling et al. 2004; Lavandier and Culling, 2010). In this theory, the binaural system uses internally generated delays to compensate for the external delay of the masker and then subtracts the stimulus at one ear from that at the other. The predictions of this equalisation-cancellation (E-C) theory are almost indistinguishable from those derived using theories based on correlation, especially in their predictions of detection thresholds. However, Culling (2007) discovered that the two theories give increasingly divergent predictions at supra-threshold signal levels necessary for the more demanding task of speech understanding. Instead of using the typical task of requiring listeners to detect a signal in noise, Culling gave listeners a loudness discrimination task. The results were entirely consistent with E-C theory.
Binaural unmasking can also be used to explain a family of illusions known as dichotic pitches, which are generated purely by the interaural phase relationships (Culling et al., 1998a,b; Culling, 1999; Culling, 2000a,b).
Recently, I have been increasingly interested in the role of reverberation on the cocktail-party problem. Culling et al. (1994) employed virtual simulations rooms to measure the effect of reverberation upon spatial unmasking and upon the DF0 effect. Culling et al. presented synthesized ‘target’ vowel sounds against ‘masker’ vowels or pink noise, and listeners were required to identify the target vowels. In anechoic conditions, spatial separation of target and masker resulted in improved vowel-identification thresholds compared to when they were co-located, but in reverberation, this unmasking effect was abolished. Lavandier and Culling (2007, 2008) showed that the effect of reverberation on binaural unmasking is mediated by reduced interaural coherence of the masker. Reverberation also affected the intrinsic intelligibility of the target, but this effect occurred only at higher levels of reverberation.
Culling et al. (1994) also found that the effect of differences in F0 between competing vowels was robust in reverberation if the F0s of the two vowels were constant, but for modulated F0s, this effect was also abolished. This observation has considerable practical significance, because natural speech always has modulation in F0 from (among other things) its intonation. We are currently investigating the effects of reverberation on both theDF0 effect and spatial unmasking through an EPSRC grant.
EPSRC (£349K) “Effects of reverberation on conversation in rooms”
Sam Jelfs (Ph.D. student, Psychology/Architecture, Cardiff University)
Dr Mike Fedeski (Architecture, Cardiff University).
Dr Tony Watkins (Psychology, University of Reading).
Dr Mathieu Lavandier (ENTPE, Université de Lyon)
Postgraduate research interests
My current research interests focus around the cocktail-party problem and binaural hearing. Specific interests include binaural unmasking, dichotic pitches, speech perception in noise, perceptual segregation by differences in F0, dip listening, temporal and spectral resolution of the binaural system, and the effects of room reverberation.
If you are interested in applying for a PhD, or for further information regarding my postgraduate research, please contact me directly (contact details available on the 'Overview' page), or submit a formal application here.
Sam Jelfs is jointly supervised by Dr Mike Fedeski from the Welsh School of Architecture. He is working on computationally efficient models of speech perception in noise in order to apply them to the acoustic design of rooms. He generates visualisations of speech intelligibility across a room – intelligibility maps. A target voice is placed a particular virtual position and masking noises at other positions, and a map is generated of the intelligibility that would be experienced at every point in the room.
Barrie Edmonds graduated with a Ph.D. in 2003. He worked on the role of across-freqeuncy processes in binaural segregation of speech in noise and against competing speech. He has since had post-doctoral positions with myself and with Dr. Katrin Krumbholz at the MRC Institute of Hearing Research. He is now a Lead Scientist at the National Biomedical Research Unit in Hearing.
Andrew Kolarik graduated with a Ph.D. in 2006. He worked on the temporal and spectral resolution of binaural unmasking. He has since had post-doctoral positions with myself and Dr Tom Freeman.
Christine Binns graduated with a Ph.D. in 2007. She worked on the role of intonation contours in speech perception in noise.
Mickael Deroche graduated with a Ph.D. in 2010. He worked on the effect of differences in F0 between competing voices on their intelligibility and the influence of room reverberation on this effect. He now has a post-doctoral position with Dr. Monita Chaterjee at the University of Maryland.
I did both my undergraduate degree in Experimental Psychology at Sussex. Following graduation I worked for several months with at the GEC Hirst Research Centre in Wembley on language processing software, before taking up a research assistantship with Dr. Ann Cutler at the MRC Applied Psychology Unit (now the Cognitive and Brain Sciences Unit). I worked there for a total of 16 months. While in Cambridge I also worked for six weeks on a phonetic labelling project in the Speech Laboratory of the Cambridge University Engineering Dept.
My doctoral thesis “The perception of double vowels” mainly concerned the effect of differences in fundamental frequency on listeners’ ability to perceptually separate concurrent speech. The experiments all involved simultaneous synthetic vowels with controlled fundamental frequency; where the vowels differed in fundamental frequency, the ability to identify both vowels in a pair was improved, suggesting that they had been perceptually separated prior to identification. The results indicated that the effect was predominantly mediated by low frequencies, but also that waveform interactions between the vowels played a substantial role. Other experiments investigate the effects of dynamic changes in fundamental frequency and in timbre.
Fellow of the Hanse Wissenschaftskolleg.
Former member of the Perceptual and Physiological Acoustics Technical Committee of the Acoustical Society of America.
1991-1995 Short-term, non-clinical scientist, MRC Institute of Hearing Research
1995-1998 MRC Research Fellow, Dept. Physiology, Oxford University and Boston University.
1998-2003 Lecturer, School of Psychology, Cardiff University.
2003-2006 Senior Lecturer, School of Psychology, Cardiff University.
2006-2009 Reader, School of Psychology, Cardiff University.
2009- Professor, School of Psychology, Cardiff University.