Hybrid Perception

This paper presents an overview of the timbre perception research. Timbre is found to be multidimensional, having a spectral envelope axis, mainly defined by the brightness of the sound, a temporal envelope, defined by the attack time, the decay slope, etc. and spectral irregularities. Furthermore, a model of quasi-harmonic sounds, the hybrid model, a timbre inspired signal model, is used to assess the just noticeable difference of a large number of perceptually relevant sound parameters. Further research is needed to better understand the perception of the parameters, and the relation to complex timbre perception, as well as to low-level perception research results. Better understanding of high-level cortex activities, such as learning and training, is also needed. Timbre perception understanding has traditionally been done using multi-dimensional scaling on the data obtained by having participants rate the differences of a large number of sounds played in pairs. These results are sometimes related to acoustic parameters. Other methods include spectral simplifications to assert the discrimination threshold of acoustic signal changes. Auditory stream segregation is also used. This involves playing a sequency of sounds, and asking subjects when two streams (or a galloping rhythm) are heard. Timbre definition seems still to be an open question, and many attempts to find words or acoustic signal parameters, which defines the values of the different timbre dimensions has been made.