Bhiksha Raj

A new layer enhances DNNs' resistance to subtle input changes.

BASS improves summarization of long audio by processing in blocks.

This article examines challenges and solutions related to noisy labels in training data.

A new method trains audio captioning systems using only text descriptions.

A new framework improves learning from incomplete data labels.

Exploring methods to improve robot performance in unpredictable environments.

New strategies enhance weak label learning by selecting relevant negative examples.

Examining how noise in pre-training data impacts model performance.

PAM offers a novel way to measure audio quality without needing reference recordings.

A new benchmark assesses voice recognition systems' performance amidst various disturbances.

Investigating how small errors in training data enhance AI-generated content.

New framework evaluates SLAM performance under challenging conditions.

New methods improve speech models for languages with limited data.

New methods improve machine understanding of human emotions in speech.

This study assesses the reasoning skills of audio-language models with a new task.

This study examines how different summarization methods affect quality and content.

A new framework enhances voice identity confirmation accuracy.

New acoustic features enhance ASR systems' performance in noisy environments.

ESPnet-Codec enhances training and evaluation of neural codecs for audio and speech.

MACE improves audio captioning by linking sounds to accurate text descriptions.

Explore how POGAT enhances the analysis of complex graph structures.

Discover how SoftVQ-VAE enhances image creation with efficiency and quality.