Traditional ASR (Signal & Cepstral Analysis, DTW, HMM) & DNNs (Custom Models + DeepSpeech) on Indian Accent Speech

Code for "Error-driven Fixed-Budget ASR Personalization for Accented Speakers" in ICASSP 2021

Multi-Teacher Knowledge Distillation for Accented English Speech Recognition

Conditioning via Adaptive Layer Norm for accented speech recognition

This repository contains the code and resources for the linguistically motivated testing framework LinguaTest, aimed at assessing ASR robustness to phonetic variations.

British English Spontaneous Speech Data