Voice assistants only understand commands thanks to complex calculations that take place in the background. There is also a great deal of mathematics involved in hearing aids: based on ambient noise, built-in software calculates in a flash which parts of this acoustic mix belong to the conversation the hearing aid user is currently having – and amplifies primarily these.
Such calculations are based on mathematical models developed, amongst others, by the Oldenburg-based hearing researcher Prof. Dr Mathias Dietz. “A model is essentially a complex formula that attempts, as accurately as possible, to describe a natural phenomenon,” he explains. In the context of his field of research, this means that the better an auditory model works, the more reliably it can predict how people would perceive a sound.
There are complex processes at work behind our intuitive sense of where a sound is coming from. “A single ear is not very good at detecting the direction of sound,” explains the researcher from the Department of Medical Physics and Acoustics. “We can only tell whether sounds are coming from the left or the right because our two ears are interconnected in the brain.” This interconnection also enables us to distinguish background noise – such as the babble of voices at a party – from the voice of the person we are talking to, and to suppress some of it without realising it.
The predecessor model is around 80 years old
One clue available to our brain when localising sound is the slight time difference with which a sound usually reaches our two ears. A sound wave coming from the left first hits the left ear and is converted into an electrical signal in the inner ear, which then travels along the auditory nerve. Because the sound wave takes longer to reach the right ear, the same process begins there with a delay of a fraction of a second. “As early as the 1940s, there was already a very intuitive idea of what happens in the brain at that moment,” says Dietz. The American physicist and psychologist Lloyd Alexander Jeffress imagined it – in simplified terms – as follows at the time:
The stimuli coming from the right and left move towards each other, passing through one nerve cell after another until they finally arrive at a single nerve cell at the same time. Because each nerve cell represents a very specific spatial direction, Jeffress hypothesised that the brain interprets the cell that is particularly strongly stimulated by two simultaneously arriving stimuli as a spatial perception; on this basis, he developed an initial model of hearing. He assumed that a large number of nerve cells were involved, acting as so-called ‘simultaneity detectors’ to map the entire environment. “This model allows us to predict sound perception quite accurately,” says Dietz. “There is just one problem: a highly branched neural structure, as Jeffress envisaged it, was not found in mammals during neuroscientific investigations some 50 years later.” Instead, they have just one nerve bundle per hemisphere – researchers refer to these as channels. The astonishing thing is that, although Jeffress started from a false assumption, his model worked – and so well, in fact, that researchers and engineers still rely on it today.
Discrepancy between physiology and models of directional hearing
More recent approaches that attempted to take physiological findings into account failed to gain acceptance. Viewing the two channels in the human brain as a system of simultaneity detectors reduced to just two nerve cells rendered the Jeffress model unusable. It did not work under this – physiologically correct – assumption. In particular, models based solely on two channels were unable to reliably predict whether people would be able to perceive target sounds when they were presented alongside background noise.
Dietz, whose research has been funded by the European Research Council (ERC) since 2018 with a prestigious ‘Starting Grant’, was already troubled by the discrepancy between physiology and the models of human directional hearing whilst working on his PhD 15 years ago. The physicist aims to understand hearing as a system. For him, this means that the findings and models contributed by various scientific disciplines must not contradict one another.
During the pandemic, when hearing experiments with test subjects were virtually impossible, Dietz, together with his colleagues Dr Jörg Encke and Bernhard Eurich, focused on finally presenting a functioning two-channel model. With success: the new Oldenburg model can reliably calculate how people perceive sounds played alongside background noise. To verify this, the researchers drew on a large number of earlier studies in which researchers had measured how loud a target sound must be, at the very least, for study participants to perceive it despite simultaneous background noise. For the first time, the Oldenburg model was able to precisely simulate more than 300 of these so-called perception thresholds.
A shift in thinking led to the breakthrough
This breakthrough, which the scientists recently published in the journal *Communications Biology*, was made possible by a shift in thinking: for the first time, the team established a relationship between the two channels. In doing so, the scientists made use of the fact that sounds travel in waves and, due to the time difference, reach each of the two ears at a different phase of this wave. The phase shift between the two channels is therefore the missing piece of the jigsaw that finally makes it possible to predict human directional hearing in a physiologically accurate manner. “We’ve really cracked a tough nut here,” says Dietz, summarising the work of recent years.
The Oldenburg approach works even better than the old model when it comes to realistically incorporating the effect of two different background noises into the prediction – something the old model had previously neglected. Eurich has outlined this in a further publication. The doctoral candidate now intends to investigate how he can use the new model to improve spatial hearing with hearing aids. The aim is to predict which parts of the soundscape must not be omitted during amplification so that a hearing aid user does not notice any loss of quality.
contact:
Publications: