We propose a physical modeling and sound synthesis model for string motion using neural networks based on modal synthesis.

We formulate a nonlinear ordinary differential equation governing the deformation of an elastomer actuator based on the hyperelastic model under dielectric stress. An explicit voltage-stretch relationship is established through numerical integration and neural network approximations.

We open-source a PyTorch-based nonlinear string sound synthesizer, which operates on a finite difference time domain.

We explore the utility of leveraging the speech quality improvement index for referenceless evaluation of acoustic echo cancellation models.

We have proposed a fast and lightweight neural network for spatial audio rendering.

We have explored a neural network-based interpolation method to address the sparsity issue in HRTF data.

We explored what an effective feature space is for anti-spoofing and investigated methods to utilize it for secure ASV.

We tried to compress/learn the numerical simulation of turbulent fluid flow using lightweight neural networks.