2017 Fiscal Year Research-status Report
Perceptual constancy in spatial hearing
| Project/Area Number |
17K12708
|
|---|
| Research Institution | Tohoku University |
|---|
Principal Investigator |
SALVADOR CESAR 東北大学, 電気通信研究所, 特任助教 (30795498)
|
|---|
| Project Period (FY) |
2017-04-01 – 2019-03-31
|
| Keywords | 音響情報処理 / 頭部伝達関数 / 位相幾何学 |
|---|
| Outline of Annual Research Achievements |
Two main goals of this research project were accomplished during the fiscal year 2017. They are as follows.
1) New mathematical methods to capture spatial sound and analyze acoustical data were formulated and evaluated. Special attention was given to distance information. Methods were formulated in abstract domains of representation where it is possible to unmask distance dependencies that are otherwise hardly observed in the natural spatial domain. This achievement constituted a step forward in the research field of spatial sound because the issues related to including distance information were not comprehensively considered in previous research. Partial results have been reported in a peer-reviewed journal paper and an international conference.
2) A collection of data describing the external anatomy of listeners' heads (morphology), as well as the transmission of sound from a point in space to the listeners' eardrums (acoustical), was created. Near sources (within 1 meter from the head) were specially considered due to their importance when aiming at rendering spatial sound in the peripersonal space with high levels of realism and naturalness. This was a major contribution because existing datasets mainly provide information for far sources (beyond 1 m from the head).
To foster the reproducibility of research results and reach the wider community, a sample of the morphological and acoustical dataset has been made publicly available at the link below:
http://www.ais.riec.tohoku.ac.jp/~salvador/download.html
|
| Current Status of Research Progress |
Current Status of Research Progress
2: Research has progressed on the whole more than it was originally planned.
Reason
The research tasks of this project are progressing as expected. In fact, the two main goals envisaged in the research plan for the fiscal year 2017 have been fulfilled: 1) new methods to capture and analyze acoustical data were formulated, and 2) a collection of morphological and acoustical datasets was created.
Regarding the first goal, previous experience on mathematical modeling in abstract domains of representation was an important point of departure to formulate the methods required by this project. As for the numerical evaluation of the proposed methods, the acquisition of a high-speed computing system allowed to reduce the computation time required by the large-scale signal processing routines associated with evaluations.
Achieving the second goal was possible thanks to a permission to use the scalar-parallel supercomputer of the Cyberscience Center of Tohoku University. A specialized application for numerical acoustics was compiled on such supercomputer. This allowed for the acoustic analysis of high-resolution head models and the creation of the collection of morphological and acoustical data in the expected time.
Finally, to foster the reproducibility and creation of new work based on the partial results of this research, special attention was given to using a standard format for storing and sharing acoustic data. The spatially oriented format for acoustics (SOFA) was adopted for his purpose. Additional efforts were also made to socialize the partial results to the wider community by means of making them publicly available on a website.
|
| Strategy for Future Research Activity |
The research plan for the fiscal year 2018 will follow according to the one proposed initially. The goal is to analyse the collection of morphological and acoustical data created during FY2017, so as to identify patterns of invariability that might potentially be correlated with perceptual constancy cues for sound localization. The plan comprises three tasks.
1) Preparation of the virtual environment for reproducible psychophysical experimentation based on headphones. This task is expected to be completed within the three first months of FY2018.
2) Design of subjective tests for auditory localization. Tests will be carried out in a comprehensive framework for experimentation. This framework comprises perceptual experiments in a controllable virtual environment, and numerical predictions based on auditory models. Experiments will be designed according to the standard protocols. This task is expected to be completed within the first quarter of FY2018.
3) Psychophysical experiments of auditory localization by presenting audio stimuli synthesized from invariant patterns.
The hypotheses of this research are expected to be confirmed by establishing correlations between patterns of invariability in the database and perceptual constancy cues. In case of not confirming the hypotheses, the contingency measure is to analyze the dataset by using conventional techniques of dimensional reduction that has proven to be promising. Such results would still contribute to the understanding of the contribution of individual features to the perceptual mechanism involved in sound localization.
|
| Causes of Carryover |
During FY2017, the main expenses required by the present project were expected to come from the use of medical services of magnetic resonance imaging to obtain 3D models of human heads. These expenses were initially considered as part of the "miscellaneous" direct costs. However, thanks to a collaboration with other researchers from Tohoku University, it was possible to use some existing high-resolution models that were obtained with the same imaging technique. For this reason, there was an available budget that was used during FY2017, and there is an incurring amount to be used during FY2018.
During FY2017, part of the "miscellaneous" budget intended to capture 3D models was used to purchase high-definition audio reproduction equipment for psychoacoustic experiments.
During FY2018, the incurring amount will be used to purchase high-definition audio recording equipment to create the contents that will be presented during the psychoacoustical experiments, and also to cover some travel expenses for international collaboration.
|
| Remarks |
Near-distance HRTFs are important to increase the realism when presenting sounds in the peripersonal space. A collection of HRTFs has been created and made publicly available as part of this project.
|
Research Products
(3 results)
