Far-Field ASR Without Parallel Data

By Vijayaditya Peddinti, Vimal Manohar, Yiming Wang, Daniel Povey, Sanjeev Khudanpur. 2016

48 citations recorded by the source corpus.

Subfields: Electrical and Electronic Engineering, Mechanical Engineering, Ocean Engineering, Speech & Audio

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Electrical and Electronic Engineering

In Mechanical Engineering

In Ocean Engineering

In Speech & Audio

Citations in the corpus

Toward Domain-Invariant Speech Recognition via Large Scale Training (2018, background): “For robustness to noise, multicondition training (MTR) using simulated noisy utterances has been shown to generalize well [3, 4].”
Purely Sequence-Trained Neural Networks for ASR Based on Lattice-Free MMI (2016, methodology): “†: In AMI LVCSR task adding a data filtering stage to eliminate large oracle-WER utterances increased the relative gain [21] * : We used a data cleanup stage to remove parts of training recordings that are mistranscribed or do not correpsond to the marked speaker.”
Recognizing Long-Form Speech Using Streaming End-to-End Models (2019, background): “Techniques such as multi-condition training have been shown to improve performance in noisy and farfield conditions [29, 30].”
Low Latency Acoustic Modeling Using Temporal Convolution and LSTMs (2018, methodology): “This recipe is same as the one described in [27], where the SDM and MDM LVCSR systems are trained with numerator lattices generated from parallel IHM data.”
The fifth 'CHiME' Speech Separation and Recognition Challenge: Dataset, task and baselines (2018, background): “This stage removes several irregular utterances, which improves the final performance of the system [38].”
A Streaming On-Device End-To-End Model Surpassing Server-Side Conventional Model Quality and Latency (2020, methodology): “3.1 and 3.2, multi-condition training (MTR) [27, 28] and random data down-sampling to 8kHz [29] are also used to further increase data diversity.”
A Study of Enhancement, Augmentation, and Autoencoder Methods for Domain Adaptation in Distant Speech Recognition (2018, background): “and has some other factors, such as cross talking [36], that need to be studied further.”
A study on data augmentation of reverberant speech for robust speech recognition (2017, methodology): “The training of baseline systems are same as [15] for the SWBD task, [4] for the ASpIRE task and [16] for the AMI task.”