Peer-reviewed research

Deep Neural Networks for Acoustic Modeling in Speech Recognition: The Shared Views of Four Research Groups

Geoffrey Hinton et al., IEEE Signal Processing Magazine, 2012 · DOI 10.1109/MSP.2012.2205597

Read the original

Read on October 17, 2026

What it is: A review article by four research groups that describes how deep neural networks replaced Gaussian mixture models in the acoustic part of speech recognition systems. It is the standard reference for the shift to neural acoustic models around 2012.
How much weight it carries: Peer-reviewed magazine article by the groups that led the work. It is a review of 2012 systems, so its benchmark results are not current. The PDF is a copy hosted by the author and marked as an IEEE proof, not the typeset issue.

Facts taken from this source

The authors write that deep neural networks with many hidden layers, trained with new methods, have been shown to outperform Gaussian mixture models on a variety of speech recognition benchmarks.
“that have many hidden layers and are trained using new methods have been shown to outperform GMMs on a variety of speech recognition benchmarks, sometimes by a large margin.” (Opening paragraph)
The paper says most speech recognition systems used hidden Markov models to handle the timing of speech, and Gaussian mixture models to score how well each HMM state fits a short window of acoustic input.
“speech recognition systems use hidden Markov models (HMMs) to deal with the temporal variability of speech and Gaussian mixture models (GMMs) to determine how well each state of each HMM fits a frame or a short window of frames of coefficients that represents the acoustic input.” (Opening paragraph)
In the 2012 Switchboard experiment, a seven-hidden-layer deep network with 2,048 units in each layer replaced the Gaussian mixture model in the acoustic model.
“A seven-hidden-layer DBN-DNN with 2,048 units in each layer and full connectivity between adjacent layers replaced the GMM in the acoustic model.” (Switchboard speech recognition task)
In a 2012 review, a deep neural network reduced the word error rate of a GMM-HMM baseline from 27.4 to 18.5 percent on a Switchboard telephone task.
“Using a DNN reduced the word error rate (WER) from the 27.4% of the baseline GMM-HMM (trained with BMMI) to 18.5%” (Switchboard speech recognition task, Table 2)

Pages that rest on this source

Peer-reviewed research

Deep Neural Networks for Acoustic Modeling in Speech Recognition: The Shared Views of Four Research Groups

Geoffrey Hinton et al., IEEE Signal Processing Magazine, 2012 · DOI 10.1109/MSP.2012.2205597

Read the original

Read on October 17, 2026

What it is: A review article by four research groups that describes how deep neural networks replaced Gaussian mixture models in the acoustic part of speech recognition systems. It is the standard reference for the shift to neural acoustic models around 2012.
How much weight it carries: Peer-reviewed magazine article by the groups that led the work. It is a review of 2012 systems, so its benchmark results are not current. The PDF is a copy hosted by the author and marked as an IEEE proof, not the typeset issue.

Facts taken from this source

The authors write that deep neural networks with many hidden layers, trained with new methods, have been shown to outperform Gaussian mixture models on a variety of speech recognition benchmarks.
“that have many hidden layers and are trained using new methods have been shown to outperform GMMs on a variety of speech recognition benchmarks, sometimes by a large margin.” (Opening paragraph)
The paper says most speech recognition systems used hidden Markov models to handle the timing of speech, and Gaussian mixture models to score how well each HMM state fits a short window of acoustic input.
“speech recognition systems use hidden Markov models (HMMs) to deal with the temporal variability of speech and Gaussian mixture models (GMMs) to determine how well each state of each HMM fits a frame or a short window of frames of coefficients that represents the acoustic input.” (Opening paragraph)
In the 2012 Switchboard experiment, a seven-hidden-layer deep network with 2,048 units in each layer replaced the Gaussian mixture model in the acoustic model.
“A seven-hidden-layer DBN-DNN with 2,048 units in each layer and full connectivity between adjacent layers replaced the GMM in the acoustic model.” (Switchboard speech recognition task)
In a 2012 review, a deep neural network reduced the word error rate of a GMM-HMM baseline from 27.4 to 18.5 percent on a Switchboard telephone task.
“Using a DNN reduced the word error rate (WER) from the 27.4% of the baseline GMM-HMM (trained with BMMI) to 18.5%” (Switchboard speech recognition task, Table 2)