Dimitri Kanevsky
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Dimitri Kanevsky has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
- Alex Olwal
Google Research, Mountain View, CA, USA
2 shared publications - Artem Dementyev
Google DeepMind, Mountain View, California, USA
2 shared publications - Chiong Lai
Google Research, Mountain View, California, USA
2 shared publications - Mathieu Parvaix
Google Research, Mountain View, California, USA
2 shared publications - Samuel J. Yang
Google Research, Mountain View, California, USA
2 shared publications - Charlotte M. Reed
Research Laboratory of Electronics, Massachusetts Institute of Technology, Cambridge, MA, USA
1 shared publication
Latest Publications
- Open Access
- 25 April 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-17
SpeechCompass: Enhancing Mobile Captioning with Diarization and Directional Guidance via Multi-Microphone Localization
- 25 April 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-17
Open AccessSpeechCompass: Enhancing Mobile Captioning with Diarization and Directional Guidance via Multi-Microphone Localization
- 1 January 2025
- Published by Institute of Mathematics, Polish Academy of Sciences in Banach Center Publications
- Vol. 129, 103-114
A universal non-associative Moufang loop connected to a cubic surface
Let $V$ be the cubic surface defined by the equation $T_0^3+T_1^3+T_2^3+\theta T_3^3=0$ over the quadratic extension $k=\mathbb {Q}_3(\theta )$ of the 3-adic rationals, where $\theta ^3=1$. We show that a relation on $V(k)$ modulo $(1-\theta )^3$ (in the- 1 January 2025
- Published by Institute of Mathematics, Polish Academy of Sciences in Banach Center Publications
- Vol. 129, 103-114
A universal non-associative Moufang loop connected to a cubic surface
Let $V$ be the cubic surface defined by the equation $T_0^3+T_1^3+T_2^3+\theta T_3^3=0$ over the quadratic extension $k=\mathbb {Q}_3(\theta )$ of the 3-adic rationals, where $\theta ^3=1$. We show that a relation on $V(k)$ modulo $(1-\theta )^3$ (in the- Open Access
- 8 November 2023
- Published by Springer Nature in Journal of Algebraic Combinatorics
- Vol. 58 (4) , 1277-1298
An example of a non-associative Moufang loop of point classes on a cubic surface
Let V be a cubic surface defined by the equation $$T_0^3+T_1^3+T_2^3+\theta T_3^3=0$$ over a quadratic extension of 3-adic numbers $$k=\mathbb {Q}_3(\theta )$$ , where $$\theta ^3=1$$ . We show that a relation on a set of geometric k-points on V modulo $$(1-\theta )^3$$ (in a ring of integers of k) defines an admissible relation and a commutative Moufang loop associated with classes of this admissible equivalence is non-associative. This answers a problem that was formulated by Yu. I. Manin more than 50 years ago about existence of a cubic surface with a non-associative Moufang loop of point classes.- 8 November 2023
- Published by Springer Nature in Journal of Algebraic Combinatorics
- Vol. 58 (4) , 1277-1298
Open AccessAn example of a non-associative Moufang loop of point classes on a cubic surface
Let V be a cubic surface defined by the equation $$T_0^3+T_1^3+T_2^3+\theta T_3^3=0$$ over a quadratic extension of 3-adic numbers $$k=\mathbb {Q}_3(\theta )$$ , where $$\theta ^3=1$$ . We show that a relation on a set of geometric k-points on V modulo $$(1-\theta )^3$$ (in a ring of integers of k) defines an admissible relation and a commutative Moufang loop associated with classes of this admissible equivalence is non-associative. This answers a problem that was formulated by Yu. I. Manin more than 50 years ago about existence of a cubic surface with a non-associative Moufang loop of point classes. - Open Access
- 29 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 1-3
LiveLocalizer: Augmenting Mobile Speech-to-Text with Microphone Arrays, Optimized Localization and Beamforming
Speech-to-text capabilities on mobile devices have proven helpful for language translation, note-taking, hearing and speech accessibility, and meeting transcripts. However, their usefulness is constrained by being unable to distinguish between multiple speakers, track which direction speech is coming from, and provide acceptable performance in noisy environments. This work introduces efficient real-time audio localization and adaptive beamforming algorithms on custom sound perception hardware running on a low-power microcontroller and four integrated microphones. A prototype is implemented in a phone case form factor and is plug-and-play with modern smartphones. We characterize the performance in technical evaluations of localization, beamforming, and diarization. We demonstrate how the phone case extends existing smartphones with speaker diarization in a speech-to-text app, sound direction visualization, and sound enhancement through beamforming. In the future, we hope our approach will inspire the widespread adoption of advanced microphone arrays that natively unlock the potential of spatial sound processing and perception in mobile and wearable devices.
- 29 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 1-3
Open AccessLiveLocalizer: Augmenting Mobile Speech-to-Text with Microphone Arrays, Optimized Localization and Beamforming
Speech-to-text capabilities on mobile devices have proven helpful for language translation, note-taking, hearing and speech accessibility, and meeting transcripts. However, their usefulness is constrained by being unable to distinguish between multiple speakers, track which direction speech is coming from, and provide acceptable performance in noisy environments. This work introduces efficient real-time audio localization and adaptive beamforming algorithms on custom sound perception hardware running on a low-power microcontroller and four integrated microphones. A prototype is implemented in a phone case form factor and is plug-and-play with modern smartphones. We characterize the performance in technical evaluations of localization, beamforming, and diarization. We demonstrate how the phone case extends existing smartphones with speaker diarization in a speech-to-text app, sound direction visualization, and sound enhancement through beamforming. In the future, we hope our approach will inspire the widespread adoption of advanced microphone arrays that natively unlock the potential of spatial sound processing and perception in mobile and wearable devices.
- Open Access
An example of a non-associative Moufang loop of point classes on a cubic surface
Let $V$ be a cubic surface defined by the equation $T_0^3+T_1^3+T_2^3+\theta T_3^3=0$ over a quadratic extension of 3-adic numbers $k=\mathbb{Q}_3(\theta)$, where $\theta^3=1$. We show that a relation on a set of geometric k-points on $V$ modulo $(1-\theta)^3$ (in a ring of integers of $k$) defines an admissible relation and a commutative Moufang loop associated with classes of this admissible equivalence is non-associative. This answers a problem that was formulated by Yu. I. Manin more than 50 years ago about existence of a cubic surface with a non-associative Moufang loop of point classes.Open AccessAn example of a non-associative Moufang loop of point classes on a cubic surface
Let $V$ be a cubic surface defined by the equation $T_0^3+T_1^3+T_2^3+\theta T_3^3=0$ over a quadratic extension of 3-adic numbers $k=\mathbb{Q}_3(\theta)$, where $\theta^3=1$. We show that a relation on a set of geometric k-points on $V$ modulo $(1-\theta)^3$ (in a ring of integers of $k$) defines an admissible relation and a commutative Moufang loop associated with classes of this admissible equivalence is non-associative. This answers a problem that was formulated by Yu. I. Manin more than 50 years ago about existence of a cubic surface with a non-associative Moufang loop of point classes. - 18 August 2022
- Published by Springer Nature
- p. 25-34
Phonemic-based Tactile Supplement to Lipreading: Initial Findings on Consonant Identification
- 18 August 2022
- Published by Springer Nature
- p. 25-34
Phonemic-based Tactile Supplement to Lipreading: Initial Findings on Consonant Identification
- Open Access
Parrotron: An End-to-End Speech-to-Speech Conversion Model and its Applications to Hearing-Impaired Speech and Speech Separation
We describe Parrotron, an end-to-end-trained speech-to-speech conversion model that maps an input spectrogram directly to another spectrogram, without utilizing any intermediate discrete representation. The network is composed of an encoder, spectrogram and phoneme decoders, followed by a vocoder to synthesize a time-domain waveform. We demonstrate that this model can be trained to normalize speech from any speaker regardless of accent, prosody, and background noise, into the voice of a single canonical target speaker with a fixed accent and consistent articulation and prosody. We further show that this normalization model can be adapted to normalize highly atypical speech from a deaf speaker, resulting in significant improvements in intelligibility and naturalness, measured via a speech recognizer and listening tests. Finally, demonstrating the utility of this model on other speech tasks, we show that the same model architecture can be trained to perform a speech separation task
Open AccessParrotron: An End-to-End Speech-to-Speech Conversion Model and its Applications to Hearing-Impaired Speech and Speech Separation
We describe Parrotron, an end-to-end-trained speech-to-speech conversion model that maps an input spectrogram directly to another spectrogram, without utilizing any intermediate discrete representation. The network is composed of an encoder, spectrogram and phoneme decoders, followed by a vocoder to synthesize a time-domain waveform. We demonstrate that this model can be trained to normalize speech from any speaker regardless of accent, prosody, and background noise, into the voice of a single canonical target speaker with a fixed accent and consistent articulation and prosody. We further show that this normalization model can be adapted to normalize highly atypical speech from a deaf speaker, resulting in significant improvements in intelligibility and naturalness, measured via a speech recognizer and listening tests. Finally, demonstrating the utility of this model on other speech tasks, we show that the same model architecture can be trained to perform a speech separation task