Statistical parametric synthesis quality surpasses knowledge-based but trails unit-selection synthesis.
Evidence
There are different motivations for building such models. First of all, being able to model the functionalities of different parts of the system provides basic knowledge of the auditory system. Often, more than one mechanism affects the perceived characteristics of an auditory object, making the analysis of the results of psychoacoustic tests complicated. The complications can be mitigated when the effects of the mechanisms can be isolated in the computational domain, making the interpretation of the results easier. Second, a very important application for such models is the evaluation of audio quality. Modern acoustic systems comprise various transfers in the acoustic, signal processing, and telecommunication domains. The measurement of quality of reproduced audio cannot be performed with simple linear measures, which calls for the usage of models of human hearing for such tasks. Ideally, the fidelity of audio systems could be estimated by computing the auditory cues reaching the conscious level. The third application of auditory models is in the diagnostics of hearing-impaired patients. The symptoms of the patient would be characterized with specific psychoacoustic tests, and the attributes of the model of human hearing were adjusted until the symptoms were explained. Optimally, this would track down the reason for the symptoms. — Ville Pulkki, Matti Karjalainen, Communication Acoustics- An…, 821 (
#expands, via linked claim)
Categorization
- XREF: connects to: Unit-selection synthesis achieves naturalness at the cost of huge databases and inflexibility; Source-filter parametric synthesis enables voice modification that unit-selection cannot. — Ville Pulkki, Matti Karjalainen, Communication Acoustics- An…, 2263; Computational auditory models serve knowledge, audio-quality measurement, and diagnostics — Ville Pulkki, Matti Karjalainen, Communication Acoustics- An…, 821
Related
- Source-filter parametric synthesis enables voice modification that unit-selection cannot. (similarity 0.87)
- Unit-selection synthesis achieves naturalness at the cost of huge databases and inflexibility (similarity 0.87)
- Data-based speech synthesis splits into sample-reassembly and parameter-model approaches. (similarity 0.85)
- Computational auditory models serve knowledge, audio-quality measurement, and diagnostics (similarity 0.84)
- Physically modeling sound sources enables flexible digital instruments (similarity 0.83)
- Digital audio growth made understanding human audition essential (similarity 0.83)
- Computational models extract quantitative metrics for the ear’s perceptual dimensions (similarity 0.81)
- Higher-level hearing models rely on assumptions tested only against psychoacoustic data (similarity 0.81)
- Computational sound quality models are trained on subjective listening data (similarity 0.80)
- Auditory models link acoustic stimulus features to perceived attributes computationally. (similarity 0.80)
- A complete mathematical model of the entire auditory system will never exist (similarity 0.80)
- Psychoacoustic models explain experimental results without addressing physiology (similarity 0.79)
- A unified source-filter model describes human and synthetic voice alike (similarity 0.79)
- Auditory modelling accuracy trades off sharply against computational cost (similarity 0.79)
- Computational auditory modelling began only in the 1960s (similarity 0.79)
Broader Connections
- Computational auditory models serve knowledge, audio-quality measurement, and diagnostics
#expands(0.75) — Ville Pulkki, Matti Karjalainen, Communication Acoustics- An…, 821