Some solution to the missing feature problem in data classification, with application to noise robust ASR

Morris, Andrew; Cooke, Martin; Green, Phil

doi:10.1109/icassp.1998.675370

Cited by 56 publications

(45 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Simple linear combination showed satisfactory performance in clean speech, while hand selection of the correct expert (when present) showed the potential of this system for very strong robustness to band limited noise. This confirmed the following result, which had previously been demonstrated in the context of missing feature theory (MFT) [25,32]: an effective strategy for reducing the effects of data mismatch is to detect and simply ignore strongly mismatching data.…”

Section: Multi-band Asr With Latent Variablessupporting

confidence: 87%

“…Conventional robust preprocessing can further increase this separation. Recognition experiments with band limited noise in both psychoacoustics [1,11] and ASR [32,25] have shown that narrow sub-bands of clean data can often be sufficient for speech recognition. There is therefore great potential for improved noise robustness with any system which is able to detect local data mismatch and focus recognition on the clean speech data which remains.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Multi-stream adaptive evidence combination for noise robust ASR

Morris

Hagen

Glotin

et al. 2001

Speech Communication

View full text Add to dashboard Cite

In this paper we develop different mathematical models in the framework of the multi-stream paradigm for noise robust ASR, and discuss their close relationship with human speech perception. Largely inspired by Fletcher's "product-of-errors" rule in psychoacoustics, multiband ASR aims for robustness to data mismatch through the exploitation of spectral redundancy, while making minimum assumptions about noise type. Previous ASR tests have shown that independent sub-band processing can lead to decreased recognition performance with clean speech. We have overcome this problem by considering every combination of data sub-bands as an independent data stream. After introducing the background to multi-band ASR, we show how this "full combination" approach can be formalised, in the context of HMM/ANN based ASR, by introducing a latent variable to specify which data sub-bands in each data frame are free from data mismatch. This enables us to decompose the posterior probability for each phoneme into a reliability weighted integral over all possible positions of clean data. This approach offers great potential for adaptation to rapidly changing and unpredictable noise.

show abstract

Section: Multi-band Asr With Latent Variablessupporting

confidence: 87%

Section: Introductionmentioning

confidence: 99%

Multi-stream adaptive evidence combination for noise robust ASR

Morris

Hagen

Glotin

et al. 2001

Speech Communication

View full text Add to dashboard Cite

show abstract

“…The missing data strategy works on the assumption that redundancy in the speech signal allows successful recognition even when moderate amounts of the signal are corrupted or obscured. Robust recognition performance in the face of missing data can be obtained, and further improvements are possible when models of auditory spectral induction (Warren et al, 1997) are incorporated (Green et al, 1995;Morris et al, 1998). In a similar vein, Berthommier et al (1998) incorporate CASAstyle information into speech recognition by varying the weights of separately processed frequency bands in a multi-band recognizer (Bourlard et al, 1996).…”

Section: Integration With Speech Recognitionmentioning

confidence: 99%

The auditory organization of speech and other sources in listeners and computational models

Cooke

Ellis

2001

Speech Communication

155

View full text Add to dashboard Cite

Speech is typically perceived against a background of other sounds. Listeners are adept at extracting target sources from the acoustic mixture reaching the ears. The auditory scene analysis (ASA) account holds that this feat is the result of a two-stage process. In the ®rst-stage, sound is decomposed into collections of fragments in several dimensions. Subsequent processes of perceptual organization reassemble these fragments, based on cues indicating common source of origin which are interpreted in the light of prior experience. In this way, the decomposed auditory scene is processed to extract coherent evidence for one or more sources. Auditory scene analysis in listeners has been studied for several decades and recent years have seen a steady accumulation of computational models of perceptual organization. The purpose of this review is to describe the evidence for the nature of auditory organization in listeners and to explore the computational models which have been motivated by such evidence. The primary focus is on speech rather than on sources such as polyphonic music or non-speech ambient backgrounds, although all these domains are equally amenable to auditory organization. The review includes a discussion of the relationship between auditory scene analysis and alternative approaches to sound source segregation. Ó

show abstract

“…There are other theoretical approaches all of which are popular research topics in pattern recognition. Many of them rely on Bayesian or other estimation techniques for extracting "class" probabilities from partial data, by integrating or averaging over missing portions of the feature space [1,2]. Such methods also include data imputation [3], and expectation maximization [4].…”

Section: A the Missing Feature Problemmentioning

confidence: 99%

An ensemble technique to handle missing data from sensors

Mohammed

Stepenosky

Polikar³

Proceedings of the 2006 IEEE Sensors Applications Symposium, 2006.

View full text Add to dashboard Cite

-Automated classification is often used in advanced systems to monitor system events. All data, and hence features from all sensors, must be present in order to make a meaningful classification. An ensemble approach, Learn ++ .MF, was recently introduced that allows classification with up to 10% of feature missing, where several classifiers are trained on random subsets of the available sensor data. Given an instance with missing features, only those classifiers trained with the available features are then used in classification. In this paper, we present a modified approach that accommodates up to 30% missing features along with the effect of varying algorithm parameters.

show abstract

Some solution to the missing feature problem in data classification, with application to noise robust ASR

Cited by 56 publications

References 6 publications

Multi-stream adaptive evidence combination for noise robust ASR

Multi-stream adaptive evidence combination for noise robust ASR

The auditory organization of speech and other sources in listeners and computational models

An ensemble technique to handle missing data from sensors

Contact Info

Product

Resources

About