Gammatone Features and Feature Combination for Large Vocabulary Speech Recognition

By Ralf Schlüter, I. Bezrukov, H. Wagner, Hermann Ney. 2007

188 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Natural Language Processing, 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 Natural Language Processing

In Speech & Audio

In Signal Processing

Citations in the corpus

RWTH ASR Systems for LibriSpeech: Hybrid vs Attention - w/o Data Augmentation (2019, methodology): “features from the triphone system and Gammatone filters [14] with 25, 50 or 100-dim.”
Multi-Task Self-Supervised Learning for Robust Speech Recognition (2020, positive): “• Adding more features: we added new workers that estimate 40 FBANKS and 40 Gammatone features [25].”
Improved training of end-to-end attention models for speech recognition (2018, methodology): “We use 40-dimensional Gammatone features [58], and the feature extraction was done with RASR [59].”
Speech acoustic modeling from raw multichannel waveforms (2015, methodology): “Such an auditory-scale gammatone filterbank has been shown to yield speech recognition performance comparable to mel-fb [16].”
Acoustic modeling with deep neural networks using raw time signal for LVCSR (2014, methodology): “Instead of the STFT based analysis, the features are extracted from an audiologically motivated filterbank realized by timedomain Gammatone filters [15].”
Advancing Sequence-to-Sequence Based Speech Recognition (2019, background): “Its input is 50-dimensional speaker independent, utterance-level mean-normalized Gammatone (GT) features [26].”
End-to-End Multimodal Emotion Recognition Using Deep Neural Networks (2017, methodology): “nels, such as gammatone filters [36], which were formulated by studying the frequency responses of the receptive fields of auditory neurons of grass-frogs, can be used as well.”
ABX-Discriminability Measures and Applications. (Mesures de discriminabilité ABX et applications) (2016, background): “MFC or PLP coefficients [18, 112] or various kinds of spectrograms [113, 114]).”
XSleepNet: Multi-View Sequential Model for Automatic Sleep Staging (2020, methodology): “One example is audio/speech in which raw audio signals [60] can be combined with its derived representations, such as mel-scale spectrogram [61] and gammatone spectrogram [62], for recognition tasks.”
LSTM Language Models for LVCSR in First-Pass Decoding and Lattice-Rescoring (2019, methodology): “All our systems use 40 dimensional Gammatone features [29].”
A comprehensive study of deep bidirectional LSTM RNNS for acoustic modeling in speech recognition (2016, methodology): “The input features are 50-dimensional VTLN-normalized Gammatone [26].”
Adieu features? End-to-end speech emotion recognition using a deep convolutional recurrent network (2016, methodology): “More complicated hand-engineered kernels, such as gammatone filters [18], which were formulated by studying the frequency responses of the receptive fields of auditory neurons of grassfrogs, can be used as well.”
Learning the speech front-end with raw waveform CLDNNs (2015, methodology): “Time-filtered gammatone features have been shown to work better with a 10th-root compression compared to a logarithmic compression [18].”
On the limit of English conversational speech recognition (2021, background): “Further, 7th root compression is applied instead of logarithmic, as in [24, 25].”
A Comparative Study on End-to-End Speech to Text Translation (2019, methodology): “We extract 40-dimensional Gammatone features [31] using the RASR [32].”