Date: Sun, 21 Nov 1999 16:58:08 -0500 (EST) From: IEEE 2000 conference Message-Id: <199911212158.QAA07550@dynamo.cs.columbia.edu> Subject: Your VR 2000 paper (077) To: phubbard@eece.unm.edu Content-Type: text Content-Length: 16177 The program committee regrets to inform you that your paper: 077 Three-Dimensional Audio Localization Using Wavelet-Domain Convolution was not accepted for IEEE VR 2000. This year, 31 papers were accepted of the 93 submitted. Having received our share of rejections, we realize how disappointing it is to have a paper turned down. If you feel the need to discuss your paper with us, we request that you please abide by two guidelines: 1) Decisions regarding acceptance/rejection are now final. 2) Please wait two weeks before contacting us, as we will both be traveling. Your reviews are appended at the end of this message. Thank you again for submitting a paper to IEEE VR 2000. We hope that you will join us this March in New Jersey! The IEEE Virtual Reality 2000 Program Co-chairs Steve Feiner Daniel Thalmann Columbia University Swiss Federal Institute of Technology vr2000@cs.columbia.edu thalmann@lig.di.epfl.ch **************************************************************************** IEEE VR '99 ========== DO NOT REMOVE THIS LINE ========== IEEE VR '99 VR200 PAPER Review Form Please email ASCII reviews to vr2000reviews@cs.columbia.edu by Friday October 29, 1999. Paper Number: 077-6 Title: Three-Dimensional Audio Localization using Wavelet-Domain Convolution Author(s): Hubbard, Umland, Pereyra, Miner, and Caudell 1. What is the main point of this paper? The paper describes an approach for creating virtual acoustic environments by synthesizing 3D sounds in the wavelet-domain. The ultimate goal is to develop and make available a real-time method for producing realistic 3D localized sounds. 2. If this is a research paper: Does it contribute to the field of VR? What is new and significant about the work? This paper provides a solid and clear discussion of the mathematics involved in the wavelet-domain convolution process. It develops a matrix formulation and shows how this can be rearranged to a format that decouples scales, although at significant expense in terms of time complexity and space required. The paper then compares different wavelet bases using audio localization of a piano music file as standard. This is done for various error thresholds that control the number of non-zero elements. All this work appears significant and at least some of it is new. The main contribution of the paper is to provide a mathematical foundation that has been investigated in detail and will offer a basis for experiments with different types of wavelets, error thresholds, audio files, and audio positioning. In addition the mathematical foundation can be used as the basis of user perceptual studies. 3. If this is a research paper: Is it likely to stimulate further work in virtual reality? Yes, it is quite likely that the paper will stimulate further research in exploring the design space for wavelet audio localization and for improving the computational efficiency of the wavelet operators. 4. Aside from its technical content, is the paper well written and organized? What needs to be improved? The paper is well-written and organized. 5. Has previous work been adequately cited? Answer yes or no. (If no, please mention missing citations here.) Yes, previous work is adequately cited. 6. Could an experienced practitioner in the field verify the paper's arguments or duplicate its results, or apply it techniques, based on the paper and its references? The paper presents a careful exposition of its methods, and an experienced practitioner should certainly be able to apply and extend them. 7. Is the paper too long or too short for its contributions? If so, how should it be condensed or expanded? The paper is at a good length. 8. Rate this paper for acceptance for presentation at IEEE VR. Provide an explanation of your rating. Rating: 5 = Definitely accept 4 = Probably accept 3 = Could go either way 2 = Probably reject 1 = Definitely reject I rate the paper 4. My main reservation is that it seems a work in progress rather than a finished piece of research. It mentions, for example, the possibility of a real-time implementation, but this will require singificant further work (and a different implementation, as the authors point out). Nevertheless, the paper presents a careful formulation and some results that should spur further work. IEEE VR '99 ========== DO NOT REMOVE THIS LINE ========== IEEE VR '99 VR200 PAPER Review Form Please email ASCII reviews to vr2000reviews@cs.columbia.edu by Friday October 29, 1999. Paper Number: 077-5 Title: Three-Dimensional Audio Localization using Wavelet-Domain Convolution Author(s): Hubbard, Umland, Pereyra, Miner, and Caudell ______________________________________________________________________ 1. What is the main point of this paper? The paper proposes HRTF convolution in the wavelet domain to create a VR sound system. As a rationale, it suggests that both synthesis and localization can be done in the same domain. ______________________________________________________________________ 2. If this is a research paper: Does it contribute to the field of VR? What is new and significant about the work? If this is an application paper: Does it address an application area in which VR techniques can make a significant impact? Does the paper quantify the benefit of the work it describes to that application area (industry, military educational, etc.)? As a research paper, it does not contribute any new concept. The paper does introduce the use of NSF of wavelets to HRTF. Therefore, perhaps it can be considered an application paper. ______________________________________________________________________ 3. If this is a research paper: Is it likely to stimulate further work in virtual reality? If this is an application paper: Does it describe techniques that are likely to be applied by others to the paper's area or to different areas? The paper uses an approach (wavelets) based on some assumptions that are not well justified. Why wavelets? ______________________________________________________________________ 4. Aside from its technical content, is the paper well written and organized? What needs to be improved? The paper is fairly easy to read. Techniques that are not new and not the main point of the paper (e.g. NSF) should be in the appendix. ______________________________________________________________________ 5. Has previous work been adequately cited? Answer yes or no. (If no, please mention missing citations here.) Other synthesis techniques in VR should be referenced. e.g. Fourier synthesis. ______________________________________________________________________ 6. Could an experienced practitioner in the field verify the paper's arguments, duplicate its results, or apply its techniques, based on the paper and its references? Yes. ______________________________________________________________________ 7. Is the paper too long or too short for its contribution? If so, how should it be condensed or expanded? The length is right for what it presents. ______________________________________________________________________ 8. Rate this paper for acceptance for presentation at IEEE VR (in past years, approximately 1 in 3 submitted papers have been accepted). Provide an explanation of your rating. Rating: 2 5 = Definitely accept 4 = Probably accept 3 = Could go either way 2 = Probably reject 1 = Definitely reject Explanation: One problem is with the motivation for the use of wavelets, even if inefficient. It all seems to hinge on the non-standard use of wavelets for synthesis. Yet no rationale is given or comparisons with other synthesis techniques. How is this system better than other comparable systems? As a research paper, it does not introduce any new concepts...just an application of a technique to a particular type of filter. ____________________________________________________________________ 9. Additional comments for the author(s): IEEE VR '99 ========== DO NOT REMOVE THIS LINE ========== IEEE VR '99 VR200 PAPER Review Form Please email ASCII reviews to vr2000reviews@cs.columbia.edu by Friday October 29, 1999. Paper Number: 077-4 Title: Three-Dimensional Audio Localization using Wavelet-Domain Convolution Author(s): Paul Hubbard, Kristin L. Umland, M.Cristina Pereyra, Nadine Miner, Thomas P. Caudell ______________________________________________________________________ 1. What is the main point of this paper? The paper describes research aimed at performing HRTF convolution in the wavelet domain in combination with synthesis methods that originate in that domain. ______________________________________________________________________ 2. If this is a research paper: Does it contribute to the field of VR? What is new and significant about the work? If this is an application paper: Does it address an application area in which VR techniques can make a significant impact? Does the paper quantify the benefit of the work it describes to that application area (industry, military educational, etc.)? The paper is intriguing. It builds on previous work on the synthesis of audio in the wavelet domain. However, rather than performing an inverse wavelet transform to get the audio into the time domain for classical HRTF filtering, this approach attempts to perform this filtering in the wavelet domain. This is a step toward a very important goal, namely one step synthesis and localization of 3D audio. ______________________________________________________________________ 3. If this is a research paper: Is it likely to stimulate further work in virtual reality? If this is an application paper: Does it describe techniques that are likely to be applied by others to the paper's area or to different areas? The paper is likely to stimulate further work and debate. ______________________________________________________________________ 4. Aside from its technical content, is the paper well written and organized? What needs to be improved? The paper is fairly well-written. I did not understand the NSF transformation between figures 2 and 3. In fairness, this may be because it is a complex subject and is well explained in the references and/or the reviewer may be obtuse. The paper could benefit by making clearer the advantages potentially realizable with this methodology over conventional methods, especially since the paper acknowledges severe increases in computational cost and memory for the new approach. Also, I believe the statement 'straightforward time-domain convolution requires O(n^2) operations' is misleading since efficient methods for HRTF rendering and even environmental effects are now in common use and have come down the cost curve thanks to advances in PC gaming. ______________________________________________________________________ 5. Has previous work been adequately cited? Answer yes or no. (If no, please mention missing citations here.) The citations seem ok. ______________________________________________________________________ 6. Could an experienced practitioner in the field verify the paper's arguments, duplicate its results, or apply its techniques, based on the paper and its references? I did not have time to look at the references. However, I think a little better detail within the paper itself would be useful. I did not understand the NSF transformation. The results in table 1 are hard to understand and could use more explanation within the text. In fairness, this paper presents complex issues and methodologies and would have to be much longer to meet the standard where other researchers could duplicate the work. ______________________________________________________________________ 7. Is the paper too long or too short for its contribution? If so, how should it be condensed or expanded? The paper should be lengthened some to clear up the issues discussed above. ______________________________________________________________________ 8. Rate this paper for acceptance for presentation at IEEE VR (in past years, approximately 1 in 3 submitted papers have been accepted). Provide an explanation of your rating. Rating: 4 5 = Definitely accept 4 = Probably accept 3 = Could go either way 2 = Probably reject 1 = Definitely reject Explanation: Although this paper has some problems and could be improved, it addresses an important area and presents an intriguing approach. I would be inclined to accept it, even in a fairly tight field, especially if some improvements can be made. ____________________________________________________________________ 9. Additional comments for the author(s): Interesting work, give us more! IEEE VR '99 =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D DO NOT REMOVE THIS LINE = =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D IEEE VR '99 VR200 PAPER Review Form Please email ASCII reviews to vr2000reviews@cs.columbia.edu by Friday October 29, 1999.=20 Paper Number: 077-3 Title: Three-Dimensional Audio Localization using Wavelet-Domain Convolution Author(s): Paul Hubbart, Kirstin L. Umland, M. Cristina Pereyra, Nadine Miner, Thomas P. Caudell ______________________________________________________________________ 1. What is the main point of this paper? A particular method for wavelet-domain convolution applied to audio ______________________________________________________________________ 2. If this is a research paper: Does it contribute to the field of VR?=20 It could do so sometimes in the future, but probably=20 not in an essential way. What is new and significant about the work?=20 Application of a wavelet-based operator to audio. The operator exploits the non-standard form of matrices. =20 If this is an application paper: Does it address an application area in which VR techniques can make a significant impact? Not in the near future Does the paper quantify the benefit of the work it describes to that=20 application area (industry, military educational, etc.)? No, it rather states that such a benefit is not yet given. ______________________________________________________________________ 3. If this is a research paper: Is it likely to stimulate further work in virtual reality? Rather not If this is an application paper: Does it describe techniques that are likely to be applied by others to the paper's area or to different areas? Not in the near future ______________________________________________________________________ 4. Aside from its technical content, is the paper well written and organized? What needs to be improved? Generally o.k. ______________________________________________________________________ 5. Has previous work been adequately cited? Answer yes or no. (If no, please mention missing citations here.) Yes, as far as this reviewer can judge. ______________________________________________________________________ 6. Could an experienced practitioner in the field verify the paper's=20 arguments, duplicate its results, or apply its techniques, based on the paper and its references?=20 Mostly, with some assistance of the authors. ______________________________________________________________________ 7. Is the paper too long or too short for its contribution? If so,=20 how should it be condensed or expanded? It is o.k. as it stands. ______________________________________________________________________ 8. Rate this paper for acceptance for presentation at IEEE VR (in past = years,=20 approximately 1 in 3 submitted papers have been accepted). Provide an explanation of your rating. Rating: 2 5 =3D Definitely accept 4 =3D Probably accept 3 =3D Could go either way 2 =3D Probably reject 1 =3D Definitely reject Explanation: Does not really meet the topic of the conference. The focus is on mathematics and/or signal-processing algorithms rather than on VR directly. ____________________________________________________________________ 9. Additional comments for the author(s):=20 Try the IEEE workshop on Signal Processing in Audio.