Krishna Subramani

I am a Ph.D. candidate in EE at UIUC, where I am advised by Prof. Paris Smaragdis. I like to do research at the intersection of signal processing and machine learning, applied to challenging problems in audio. Prior to this, I graduated from IIT Bombay with a B.Tech and M.Tech in Electrical Engineering.

I've had the good fortune to work with and learn from some amazing research groups: AWS Audio Science, MTG Barcelona, DAP Lab and Honda Research.

Email  /  CV  /  LinkedIn  /  GitHub

Krishna Face

  • [September 2022]   I'll be in-person at Interspeech! Do drop by our poster on Monday between 14.30 and 16.30 KST
  • [June 2022]   Our work on end-to-end LPCNet was accepted at Interspeech 2022!
  • [May 2022]   I'll be returning to AWS, Palo Alto as an Applied Scientist intern over the summer
  • [December 2021]   I passed the ECE PhD Qualifying Exam, I'm now officially a PhD Candidate!
  • [October 2021]   Point Cloud Audio Processing won the Best Paper Award at WASPAA 2021!!
  • [August 2021]   Our work on Point Cloud Audio Processing was accepted!! We'll be presenting it at WASPAA 2021
  • [May 2021]   Interned at AWS as an Applied Scientist over the summer!
  • [May 2021]   Submitted our work on Point Cloud based Audio Processing to WASPAA 2021
  • [February 2021]   Our work on a differentiable STFT pipeline was accepted!! We'll be presenting it at ICASSP 2021
  • [October 2020]   Submitted our work on a differentiable STFT pipeline to ICASSP 2021
  • [August 2020]   Graduated from IIT Bombay, beginning my Ph.D. in UIUC with Paris Smaragdis



End-to-end LPCNet: A Neural Vocoder With Fully-Differentiable LPC Estimation
Krishna Subramani, Jean-Marc Valin, Umut Isik, Paris Smaragdis, Arvindh Krishnaswamy
Interspeech 2022
Paper / Code / BibTeX

We propose an end-to-end version of LPCNet that learns to infer the LP coefficients (without explicit analysis). Our open-source end-to-end model still benefits from LPCNet's low complexity, while allowing for any type of conditioning features.

We'll be presenting our poster in-person at Interspeech 2022

Point Cloud Audio Processing
Krishna Subramani, Paris Smaragdis
WASPAA 2021 (Best Paper Award)
Paper / Code / DOI / BibTeX

We introduce a novel way of processing audio signals by treating them as a collection of points in feature space, and we use point cloud machine learning models that give us invariance to the choice of representation parameters, such as DFT size or the sampling rate

We presented our work as a virtual presentation.

Optimizing Short-Time Fourier Transform Parameters via Gradient Descent
An Zhao, Krishna Subramani, Paris Smaragdis
Paper / Code / DOI / BibTeX

We show an approach that allows us to obtain a gradient for STFT parameters with respect to arbitrary cost functions, and thus allows the use of gradient descent based optimization of quantities like the STFT window length, or the STFT hop size.

We presented our work as a virtual presentation.

VaPar Synth - A Variational Parametric Model for Audio Synthesis
Krishna Subramani, Preeti Rao Alexandre D'Hooge
Paper / Code / Audio Examples / DOI / BibTeX

We present VaPar Synth - a Variational Parametric Synthesizer for instrument note synthesis, which utilizes a conditional variational autoencoder trained on a source-filter inspired parametric representation.

We presented our work as a virtual presentation.

Energy-Weighted Multi-Band Novelty Functions for Onset Detection in Piano Music
Krishna Subramani, Srivatsan Sridhar, Rohit M A, Preeti Rao
National Conference on Communications 2018
Paper / DOI / BibTeX

Propose the use of energy-based weighting of multi-band onset detection functions and the use of a new criterion for adapting the final peak-picking threshold to improve detection of soft onsets in the vicinity of loud notes. Also propose a grouping algorithm to reduce the detection of spurious onsets.

We presented our work as an oral presentation

Generative Audio Synthesis with a Parametric Model
Krishna Subramani, Alexandre D'Hooge, Preeti Rao
ISMIR 2019 Late Breaking/Demo
Abstract / Audio Examples / BibTeX

Propose a parametric representation for audio corresponding more directly to its musical attributes such as pitch, dynamics and timbre. For more control over generation, we also propose the use of a conditional variational autoencoder which conditions the timbre on pitch.

We presented our work as a poster in the Late Breaking/Demo session


Learning Complex Representations from Spatial Phase Statistics of Natural Scenes
HaDi Maboudi, Krishna Subramani, Hamid Soltanian-Zadeh, Shun-ichi Amari, Hideaki Shimazaki
bioRxiv / BibTeX

We introduce a generative model for phase in visual systems, and propose a complex domain based maximum likelihood estimation procedure for parameter estimation. We derive analytical gradient expressions for maximum likelihood estimation using Wirtinger Calculus (detailed in our supplementary material)

I presented the initial part of this work as an oral presentation at Honda Research Institute, Saitama, Japan

HpRNet : Incorporating Residual Noise Modeling for Violin in a Variational Parametric Synthesizer
Krishna Subramani, Preeti Rao
arXiv / Code / Audio Examples / BibTeX

We investigate a parametric model for violin tones, in particular the generative modeling of the residual bow noise to make for more natural tone quality. To aid in our analysis, we record a dataset of Carnatic Violin Recordings.

We also present a simple GUI (inspired from SMS-Tools!) for researchers to play around with, where they can load pre-trained network weights and reconstruct/generate user input audio files.

Variational Parametric Models for Audio Synthesis
Thesis / Presentation / BibTeX

My Master's Thesis on generative audio synthesis, where we introduce a Variational Parametric Framework for the synthesis of instrumental tones.

I received the Undergraduate Research Award (URA03) from IIT Bombay for my thesis as recognition of truly exceptional work and research contributions.

The Master Yoda to us Padawans.