Hidden Markov model decomposition of speech and noise

Varga, Attila Károly; Moore, Roger K.

doi:10.1109/icassp.1990.115970

Cited by 344 publications

(143 citation statements)

References 4 publications

Supporting

Mentioning

140

Contrasting

Unclassified

Order By: Relevance

“…We denote the observed speech and additive noise mixture in channel d of time frame τ in the log-mel-spectral domain by Y (τ, d). According to the so-called logmax approximation (Nádas et al, 1989;Varga and Moore, 1990) the log-magnitude-compressed observations can be approximated as…”

Section: Mask Estimationmentioning

confidence: 99%

Comparing human and automatic speech recognition in a perceptual restoration experiment

Remes

López

Juvela

et al. 2016

Computer Speech & Language

View full text Add to dashboard Cite

Speech that has been distorted by introducing spectral or temporal gaps is still perceived as continuous and complete by human listeners, so long as the gaps are filled with additive noise of sufficient intensity. When such perceptual restoration occurs, the speech is also more intelligible compared to the case in which noise has not been added in the gaps. This observation has motivated so-called 'missing data' systems for automatic speech recognition (ASR), but there have been few attempts to determine whether such systems are a good model of perceptual restoration in human listeners. Accordingly, the current paper evaluates missing data ASR in a perceptual restoration task. We evaluated two systems that use a new approach to bounded marginalisation in the cepstral domain, and a bounded conditional mean imputation method. Both methods model available speech information as a clean-speech posterior distribution that is subsequently passed to an ASR system. The proposed missing data ASR systems were evaluated using distorted speech, in which spectro-temporal gaps were optionally filled with additive noise. Speech recognition performance of the proposed systems was compared against a baseline ASR system, and with human speech recognition performance on the same task. We conclude that missing data methods improve speech recognition performance in a manner that is consistent with perceptual restoration in human listeners.

show abstract

Section: Mask Estimationmentioning

confidence: 99%

Comparing human and automatic speech recognition in a perceptual restoration experiment

Remes

López

Juvela

et al. 2016

Computer Speech & Language

View full text Add to dashboard Cite

show abstract

“…This is reminiscent of the 'HMM decomposition' approach [Moore 1986, Varga & Moore 1990] which treats a mixture as the combined output of several processes, each described by a hidden Markov model. In that case, the observation probability is…”

Section: Recognizing Speech Given Nonspeech Estimatesmentioning

confidence: 99%

Using knowledge to organize sound: The prediction-driven approach to computational auditory scene analysis and its application to speech/nonspeech mixtures

Ellis

1999

Speech Communication

View full text Add to dashboard Cite

Computational auditory scene analysis -modeling the human ability to organize sound mixtures according to their sources -has experienced a rapid evolution as the simple principles suggested by psychological experiments have turned out to be less than the whole story. Phenomena such as the continuity illusion and phonemic restoration show that the brain is able to use a wide range of knowledge-based contextual constraints when interpreting obscured or complex mixtures: To model such processing, we need architectures that operate by confirming hypotheses about the observations rather than relying on directly-extracted descriptions. One such architecture, the 'prediction-driven' approach, is presented along with results from its initial implementation. This architecture can be extended to take advantage of the high-level knowledge implicit in today's speech recognizers by modifying a recognizer to act as one of the 'component models' which provide the explanations of the signal mixture. Although this adaptation raises a number of issues, a preliminary investigation supports the argument that successful scene analysis must exploit such abstract knowledge at every level. IntroductionThe work described in this paper fits into a kind of evolutionary tale of approaches to sound organization: In the beginning, there was the 'simplistic' or "blank background" view that sound objects somehow defined themselves, and that identifying a single perceptual object was as simple as picking out a figure in a child's coloring book. The experimental stimuli on which so much of our understanding of auditory organization is based -the sinusoids and bandlimited noise bursts of Bregman [1990] and others -echo this approach, since, as presented in soundproof listening booths, they would actually be amenable to such an approach.The second stage of evolution, which we might call the 'optimistic' or "uniform background" view, emerged from the initial efforts to apply the insights of experimental results in auditory organization (especially those in [Bregman 1990]) and apply them to real sounds.Unlike sinusoids against a silent background, real sounds contain all kinds of noise and distractions to defeat simple extraction routines, and therefore demand a more sophisticated approach. However, the signal processing community is long used to dealing with noise and offers various approaches for making the best possible decisions under some simple, but useful, assumptions. These amount to a kind of template matching, such that if the form of the target and the interference can be exactly specified, the parameters of the target can be recovered in the mathematically best-possible fashion. The essence of this approach is that we can produce simple definitions of what we are looking for -sinusoids of unknown frequency, or narrowband noise energy -and we can then go through a given signal identifying and extracting just the parts that interest us, and ignoring the rest -in analogy to the way a human listener is able to 'screen out' interferin...

show abstract

“…At the hidden layer, articulatory-based approaches to speech recognition are becoming more popular [6], where the speech signal is represented by multiple semi-synchronous streams of articulatory gestures. Early multistream work also includes that of HMM decomposition [7], where both speech and noise are consider a separate stream. Dynamic Bayesian networks (DBNs) have also been used for multi-stream [8], including audio-visual speech recognition [9,10,11].…”

Section: Introductionmentioning

confidence: 99%

Polyphase speech recognition

Lin

Bilmes

2008

2008 IEEE International Conference on Acoustics, Speech and Signal Processing

View full text Add to dashboard Cite

We propose a model for speech recognition that consists of multiple semi-synchronized recognizers operating on a polyphase decomposition of standard speech features. Specifically, we consider multiple out-of-phase downsampled speech features as separate streams which are modeled separately at the lowest level, and are then integrated at the higher level (words) during first-pass decoding. Our model lessens the severity of the oversampling problem in many speech recognition systems -i.e., that speech modulation energy is most important below 25Hz but a 100Hz frame rate gives a modulation bandwidth of 50Hz. Our polyphase approach moreover captures wider and more diverse dynamics within the speech signal. Our integrative network is high-level, namely it couples together and decodes word strings from different recognizers simultaneously and asynchronously. We provide preliminary results on the 10-word vocabulary version of the SVitchboard (small-vocabulary switchboard) task and show that our polyphase recognition system significantly outperforms an optimized baseline (HMM) approach.

show abstract

Hidden Markov model decomposition of speech and noise

Cited by 344 publications

References 4 publications

Comparing human and automatic speech recognition in a perceptual restoration experiment

Comparing human and automatic speech recognition in a perceptual restoration experiment

Using knowledge to organize sound: The prediction-driven approach to computational auditory scene analysis and its application to speech/nonspeech mixtures

Polyphase speech recognition

Contact Info

Product

Resources

About