Pitch is a fundamental perceptual attribute of communication sounds and music, and is a powerful cue for sorting out the different sounds in a mixture.
To what extent
does pitch perception and our system of music pitches
arise from experience of speech sounds?
Sounds are usually heard, not in isolation, but in sequences and as part of higher-level cognitive structures, and our perception of pitch is influenced by context; e.g. facilitated in the presence of a consistent melodic sequence but impaired by an inconsistent one. But...
What does consistent mean? How does it depend on previous experience?
Current models
of pitch perception generally extract absolute pitch,
but people are usually much better at categorising pitch relationships.
Are intervals represented as composite patterns, or does the auditory
system separately extract absolute pitches, before deciding upon their
relationships?
We are building an 'active' model of relative pitch perception to investigate the:
A general understanding of pitch perception is an essential element in explaining the perception of auditory scenes. Pitch is a salient percept derived from the periodicities in a sound signal; and the perception of the pitch depends on its temporal context, as it happens for example in musical and in speech. However, studies of the temporal dynamics of pitch have demonstrated that it is necessary to invoke a wide range of time scales over which perceptual information is integrated by the auditory system, and furthermore, that the appropriate analysis time scale is stimulus-dependent. This multi-scale phenomenology is not exclusive to pitch perception; it appears to be a more general characteristic of the auditory processing, and it is known as the temporal resolution-temporal integration dilemma. No existing model can account for such stimulus-dependent variations in the perceptual integration window. Our studies have recently demonstrated that feedback connections from central to pre-cortical areas may be critical for the temporal processing underlying auditory perception. This novel contribution formed the basis of a unitary compact model, which accounts for the balance between resolution and integration in auditory perception. In this model, interacting neural ensembles provide a unified account of a range of perceptual experiments and explain some of the neuromagnetic responses to pitch in cortex not previously possible in a single model.
A parallel problem is the simultaneous perception of more than a single pitch, the perceptual segregation. We have recently shown that this phenomenon can be explained in terms of the information conveyed in each of the cochlear frequency channels: the auditory nerve patterns of activity in some of the channels have similar dynamical properties; therefore, homogeneous groups of channels naturally emerge in the auditory peripheral processing. After that, we demonstrated that each group is responsible for each one of the different pitches that can be simultaneously perceived. Using this model, we have explained a wide range of perceptual segregation phenomenology not accounted for by previous approaches.
Implicit learning of music may involve the learning of statistical regularities in musical sequences or patterns. This is referred to as (short term) statistical learning in the context of our study study. We have concluded that:
The main implications of this finding is that machine learning of music (e.g., for a computer model) needs to take into account enculturation and inherent perceptual biases of the auditory system. Purely data-driven approaches are not realistic.
The video presentation "Experiments in Statistical Learning of Short Tone Sequences" by Alicia Knast and Eduardo R. Miranda describes in detail how to determine how listeners learn sequences of tones.