The aurora experimental framework for the performance evaluation of speech recognition systems under noisy conditions
By D. Pearce, H. Hirsch. 2000
2,150 citations recorded by the source corpus.
This paper describes a database designed to evaluate the performance of speech recognition algorithms in noisy conditions. The database may either be used for the evaluation of front-end feature extraction algorithms using a defined HMM recognition back-end or complete recognition systems. The source speech for this database is the TIdigits, consisting of connected digits task spoken by American English talkers (downsampled to 8kHz) . A selection of 8 different real-world noises have been added to the speech over a range of signal to noise ratios and special care has been taken to control the filtering of both the speech and noise. The framework was prepared as a contribution to the ETSI STQ-AURORA DSR Working Group [1]. Aurora is developing standards for Distributed Speech Recognition (DSR) where the speech analysis is done in the telecommunication terminal and the recognition at a central location in the telecom network. The framework is currently being used to evaluate alternative proposals for front-end feature extraction. The database has been made publicly available through ELRA so that other speech researchers can evaluate and compare the performance of noise robust algorithms. Recognition results are presented for the first standard DSR feature extraction scheme that is based on a cepstral analysis.
Subfields: Artificial Intelligence, Evaluation & Benchmarks, Speech & Audio, Signal Processing
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
In Evaluation & Benchmarks
- Shinji Watanabe
- Ronan Collobert
- Jinyu Li
- Björn Schuller
- Quoc V. Le
- Emmanuel Vincent
- Jonathan Le Roux
- Yifan Gong
In Speech & Audio
In Signal Processing
Citations in the corpus
- Analysis of CNN-based speech recognition system using raw speech as input (2015, background): “The Aurora2 corpus [24] is a connected digit corpus which contains 8,440 sentences of clean and multi-condition training data and 70,070 sentences of clean and noisy test data.”
- Deep Belief Networks Based Voice Activity Detection (2013, methodology): “1) Dataset: Seven noisy test corpora of AURORA2 [45] is used for performance analysis.”
- Deep and Wide: Multiple Layers in Automatic Speech Recognition (2012, background): “In 2000, as part of a European Telecommunications Standards Institute (ETSI) competition for a new Distributed Speech Recognition standard [11], an approach to speech recognition was developed that was called Tandem [12].”
- Comparing multilayer perceptron to Deep Belief Network Tandem features for robust ASR (2011, methodology): “For this paper, we use the Aurora2 data set described in [10], a connected digit corpus which contains 8,440 sentences of clean training data and 56,056 sentences of clean and noisy test data.”
- Benchmarking Neural Network Robustness to Common Corruptions and Surface Variations (2018, background): “There are several popular datasets containing noisy test audio [25, 24].”
- The second ‘chime’ speech separation and recognition challenge: Datasets, tasks and baselines (2013, background): “In comparison to conventional robust ASR evaluations [5] which operate by rescaling the speech and noise signals, this mixing procedure is ecologically more valid, although it does not yet account for the Lombard effect as in [6, 7].”
- An efficient MFCC extraction method in speech recognition (2006, methodology): “The speech signals used in this paper are all from the Aurora 2 database [7].”
- Speech enhancement with missing data techniques using recurrent neural networks (2004, methodology): “The second stage chooses the regions of noisy speech with higher levels of corruption by using a missing data mask, , and replaces them with the enhanced speech, , coming from RNN enhancer.”
- Robust Speech Recognition Using a Cepstral Minimum-Mean-Square-Error-Motivated Noise Suppressor (2008, methodology): “The Aurora-3 task consists of noisy digit recognition subtasks under realistic automobile environments [10], [15].”