deep neural networks for acoustic modeling in speech recognition

Introduction

Classical speech recognition systems use hidden Markov models (HMMs) to model the temporal variations and Gaussian mixture models (GMMs) to determine the likelihood of each state of each HMM given an acoustic observation. The speech signal in the classical systems are represented by a series of Mel-frequency cepstral coefficients (MFCCs) or perceptual linear predictive coefficients (PLPs) extracted from overlapping short windows of the raw speech signal. Although GMMs are quite flexible and are fairly easy to train using the Expectation Maximization (EM) algorithm, they are inefficient when modelling data that lie close to a nonlinear manifold which is the case for the speech data. Deep Neural Networks (DNNs) don't suffer from the same shortcoming, hence they can learn much better models than GMMs. Over the past few years, training DNNs has become possible thanks to the advancements in machine learning and computer hardware, which makes it possible to replace GMMs with DNNs in the speech recognition systems. DNNs are proved to outperform GMMs in both small and large vocabulary speech recognition tasks.

Interfacing a DNN with an HMM

HMM model requires the likelihoods of the observations [math]\displaystyle{ p(AcousticInput|HMMstate) }[/math] for running the forward-backward algorithm or for computing a Viterbi alignment. DNNs output the posteriors [math]\displaystyle{ p(HMMstate|AcousticInput) }[/math] which can be converted to scaled version of the likelihood by dividing them by [math]\displaystyle{ p(HMMstate) }[/math], where [math]\displaystyle{ p(HMMstate) }[/math] is the HMM states frequencies in the training data. The conversion from the posteriors to the likelihoods is important when the training labels are highly unbalanced.

Phonetic Classification and Recognition on TIMIT

TIMIT is an acoustic-phonetic countinuous speech corpus that has been widely used as a benchmark data set for the speech recognition systems. DNN-HMM systems outperformed the classical GMM-HMM systems. The first successful attempt for building a DNN-HMM speech recognition system was published in 2009 by Mohamed et. al.<ref name=firstDBN> A. Mohamed, G. Dahl, and G. Hinton, “Deep belief networks for phone recognition,” in Proc. NIPS Workshop Deep Learning for Speech Recognition and Related Applications, 2009. </ref>, they reported a significant improvement in the accuracy over the state-of-the-art DNN-HMM systems at that time. It was found that the structure of the DNN (i.e. number of hidden layers, and number of hidden units per layer) has little effect on the accuracy, which made it possible to focus more on learning the metaparameters of the DNN. Details of the learning rates, stopping criteria, momentum, L2 weight penalties and minibatch size, pretraining, and fine-tuning can be found in <ref name=tuning_fb_DBN>A. Mohamed, G. Dahl, and G. Hinton, “Acoustic modeling using deep belief networks,” IEEE Trans. Audio Speech Lang. Processing, vol. 20, no. 1, pp. 14–22, Jan. 2012.</ref>.

Using Filter-Bank Features

MFCC features are commonly used in the GMM-HMM systems because they provide uncorrelated features, which is important to avoid using full covariance GMMs. Some of the acoustic information is lost due to using MFCCs. DNNs on the other hand can work with correlated features which opened the room for using filter-bank features. It was found that using filter-bank features with DNNs improved the accuracy by 1.7% <ref name=tuning_fb_DBN></ref>.

Fine-Tuning DNNs To Optimize Mutual Information

In the experiments mentioned earlier this section, the system were tuned to optimize the per frame cross entropy or the log posterior probablity [math]\displaystyle{ p(l_t|v_t) }[/math]; where [math]\displaystyle{ l_t }[/math] is label at time [math]\displaystyle{ t }[/math], and [math]\displaystyle{ v_t }[/math] is the features at the same time step. The transition probabilities and the language models were tuned independently using the HMM framework. The DNN can be tuned to optimize the conditional probability [math]\displaystyle{ p(l_{1:T}|v_{1:T}) }[/math], this is done for the softmax layer only and by fixing the parameters of the hidden layers [math]\displaystyle{ h }[/math].

[math]\displaystyle{ (l_{1:T}|v_{1:T}) = p(l_{1:T}|h_{1:T}) = \frac{\exp(\sum_{t=1}^T\gamma_{ij} \phi_{ij}(l_{t-1},l_t) + \sum_{t=1}^T\sum_{d=1}^D \lambda_{l_t,d} h_{td})}{Z(h_{1:T})} }[/math]

Where [math]\displaystyle{ \phi_{i,j}(l_{t-1},l_t) }[/math] is the transition feature and it takes a value of one if [math]\displaystyle{ l_{t-1} = i }[/math] and [math]\displaystyle{ l_{t} = j }[/math] and zero otherwise, [math]\displaystyle{ \gamma_{ij} }[/math] is the parameter associated with the transition feature, [math]\displaystyle{ \lambda }[/math] are the weights of the softmax layer. [math]\displaystyle{ \gamma,\lambda }[/math] are tuning using gradient descent, and the experiments show the fine tuning DNNs to optimize the mutual information improved the accuracy by 5% relative on TIMIT data set <ref name=finetuningDNN> A. Mohamed, D. Yu, and L. Deng, “Investigation of full-sequence training of deep belief networks for speech recognition,” in Proc. Interspeech, 2010, pp. 2846–2849. </ref>.

DNNs and GMMs

As you can see in the following table, monophone DNNs can outperform the best triphone GMM-HMM by 1.7%, the reason behind such success can be due to the following:

1. DNNs are instance of product of experts in which each parameter is constrained by large amount of the data, while GMMs are sum of experts for which each parameter applies to a small amount of the data.
2. GMM assumes that each datapoint is generated from a single component which makes it inefficient when it comes to modeling multiple simultaneous events. DNNs is flexible enough to model multiple simultaneous events.
3. GMMs are restricted to uncorrelated features, while GMMs can work with correlated features. This allows GMMs to use correlated features such as filter-banks, it also allows to analyse a larger window of the signal at each timestep.

Compariosons among the reported speaker-independent (SI) phonetic error rate (PER) results on TIMIT core test set with 192 sentences.

Method	PER
CD-HMM <ref name=cdhmm> Y. Hifny and S. Renals, “Speech recognition using augmented conditional random fields,” IEEE Trans. Audio Speech Lang. Processing, vol. 17, no. 2, pp. 354–365, 2009. </ref>	27.3%
Augmented Conditional Random Fienlds <ref name=cdhmm></ref>	26.6%
Randomly Initialized Recurrent Neural Nets <ref name=rirnn> A. Robinson, “An application to recurrent nets to phone probability estimation,” IEEE Trans. Neural Networks, vol. 5, no. 2, pp. 298–305, 1994. </ref>	26.1%
Bayesian Triphone GMM-HMM <ref name=btgmmhmm> J. Ming and F. J. Smith, “Improved phone recognition using Bayesian triphone models,” in Proc. ICASSP, 1998, pp. 409–412. </ref>	25.6%
Monophone HTMs <ref name=mhtms> L. Deng and D. Yu, “Use of differential cepstra as acoustic features in hidden trajectory modelling for phonetic recognition,” in Proc. ICASSP, 2007, pp. 445–448. </ref>	24.8%
Heterogeneous Classifiers <ref name=hclass> A. Halberstadt and J. Glass, “Heterogeneous measurements and multiple classifiers for speech recognition,” in Proc. ICSLP, 1998. </ref>	24.4%
Monophone Randomly Initialized DNNs (Six Layers) <ref name=tuning_fb_DBN></ref>	23.4%
Monophone DBN-DNNs (Six Layers) <ref name=tuning_fb_DBN></ref>	22.4%
Monophone DBN-DNNs with MMI Training <ref name=finetuningDNN></ref>	22.1%
Triphone GMM-HMMs DT W/BMMI <ref name=tgmmhmmbmmi> T. N. Sainath, B. Ramabhadran, M. Picheny, D. Nahamoo, and D. Kanevsky, “Exemplar-based sparse representation features: From TIMIT to LVCSR,” IEEE Trans. Audio Speech Lang. Processing, vol. 19, no. 8, pp. 2598–2613, Nov. 2011. </ref>	21.7%
Monophone DBN-DNNs on FBank (Eight Layers) <ref name=tuning_fb_DBN></ref>	20.7%
Monophone MCRBM-DBN-DNNs On FBank (Five Layers) <ref name=mmcrbmdbndnn> G. E. Dahl, M. Ranzato, A. Mohamed, and G. E. Hinton, “Phone recognition with the mean-covariance restricted Boltzmann machine,” in Advances in Neural Information Processing Systems 23, J. Lafferty, C. K. I. Williams, J. Shawe- Taylor, R.S. Zemel, and A. Culotta, Eds. Cambridge, MA: MIT Press, 2010, pp. 469–477. </ref>	20.5%
Monphone Convolutional DNNs On Fbank (Three Layers) <ref name=cdnnfb> O. Abdel-Hamid, A. Mohamed, H. Jiang, and G. Penn, “Applying convolutional neural networks concepts to hybrid NN-HMM model for speech recognition,” in Proc. ICASSP, 2012, pp. 4277–4280. </ref>	20.0%

DNN for Large-Vocabulary Speech Recognition

The success of DNN-HMM system on TIMIT data set opened the door for trying the same technique on larger data sets. It was found that using context dependent HMM states is essential for large data sets. DNN-HMM systems were tested on five different large tasks and they outperform GMM-HMM systems on every task.

Bing-Voice-Search Speech Recognition Task

Bing-Voice-Search is a 24 hrs of speech data with different sources of acoustic variations such as noise, music, side-speech, accents, sloppy pronunciations, interruptions, and mobile phone differences. The DNN-HMM system was trained based on the DNN that worked well for TIMIT. it contained 5 hidden layers of size 2048 each. A window of 11 frames was used to classify the middle frame into the corresponding HMM state, and tri-phone states were used instead of monophones. The DNN-HMM system achieved a sentence accuracy of 69% compared to 63.8% for the GMM-HMM system. The accuracy for the DNN-HMM system was further improved by increasing the data size to 48 hrs achieving a sentence accuracy of 71.7% <ref name=bing> G. Dahl, D. Yu, L. Deng, and A. Acero, “Context-dependent pretrained deep neural networks for large-vocabulary speech recognition,” IEEE Trans. Audio Speech Lang. Processing, vol. 20, no. 1, pp. 30–42, Jan. 2012. </ref>.

Switchboard Speech Recognition Task

This data set contains over 300 hrs of speech data, and the test set is 6.3 hrs of speech. The same DNN-HMM system developed for Bing data set was applied to this data set. The DNN used contains 7 hidden layers of size 2048 each. A trigram language model was used, it was trained using 2000 hrs of speech corpus. This data set is publicly available which allows for rigorous comparisons among different techniques to be performed. As shown in the table below, DNN reduced the word error rate from 27.4% to 18.5%. DNN system performed as well as a GMM system that combines several speaker-adaptive multipass systems and uses nearly 7 times as much acoustic data for training as the DNN system.

Google Voice Input Speech Recognition Task

The data set contains search engine queries, short messages, events, and user actions from mobile devices. A well-trained GMM-HMM system was used to align 5870 hrs of speech for the DNN-HMM system. The DNN used is 4 hidden layer with 2560 hidden units per layer. A window of 11 frames was used to classify the middle frame, and 40 log filter-bank features were used to represent each frame. The network was fine-tuned to maximize the mutual information. The DNN was sparsified by setting weights that are less that certain threshold to zero. To further improve the accuracy, the DNN-HMM and the GMM-HMM models were combined using segmental conditional random field framework <ref name=scrf> N. Jaitly, P. Nguyen, A. Senior, and V. Vanhoucke, “An application of pretrained deep neural networks to large vocabulary speech recognition,” submitted for publication. </ref>.

References

<references />