Adaptation to new microphones using tied-mixture normalization

Anastasakos, A.; Kubala, Francis; Makhoul, John; Schwartz, Richard

doi:10.3115/1075812.1075888

Cited by 4 publications

(3 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Speaker normalization strategies aim at reducing inter-speaker variability and thus reducing the fragmentation of the acoustic models learned by an ASR system. In general, normalization can be achieved by either explicitly removing some acoustic peculiarities of the speaker (e.g., through Cepstral mean removal, Anastasakos et al, 1994, or vocal tract length normalization, Eide and Gish, 1996), by explicitly mapping the speech acoustics of some speakers into the acoustic domain of a reference speaker (e.g., Huang, 1992), or by creating compact models (i.e., models that are robust across inter-speaker variations) that can then be adapted to the different speakers (this is usually referred to as adaptive training strategy, Anastasakos et al, 1996). …”

Section: Introductionmentioning

confidence: 99%

Modeling speech imitation and ecological learning of auditory-motor maps

Canevari¹,

Badino²,

D’Ausilio³

et al. 2013

Front. Psychol.

View full text Add to dashboard Cite

Classical models of speech consider an antero-posterior distinction between perceptive and productive functions. However, the selective alteration of neural activity in speech motor centers, via transcranial magnetic stimulation, was shown to affect speech discrimination. On the automatic speech recognition (ASR) side, the recognition systems have classically relied solely on acoustic data, achieving rather good performance in optimal listening conditions. The main limitations of current ASR are mainly evident in the realistic use of such systems. These limitations can be partly reduced by using normalization strategies that minimize inter-speaker variability by either explicitly removing speakers' peculiarities or adapting different speakers to a reference model. In this paper we aim at modeling a motor-based imitation learning mechanism in ASR. We tested the utility of a speaker normalization strategy that uses motor representations of speech and compare it with strategies that ignore the motor domain. Specifically, we first trained a regressor through state-of-the-art machine learning techniques to build an auditory-motor mapping, in a sense mimicking a human learner that tries to reproduce utterances produced by other speakers. This auditory-motor mapping maps the speech acoustics of a speaker into the motor plans of a reference speaker. Since, during recognition, only speech acoustics are available, the mapping is necessary to “recover” motor information. Subsequently, in a phone classification task, we tested the system on either one of the speakers that was used during training or a new one. Results show that in both cases the motor-based speaker normalization strategy slightly but significantly outperforms all other strategies where only acoustics is taken into account.

show abstract

Section: Introductionmentioning

confidence: 99%

Modeling speech imitation and ecological learning of auditory-motor maps

Canevari¹,

Badino²,

D’Ausilio³

et al. 2013

Front. Psychol.

View full text Add to dashboard Cite

show abstract

“…In the dry-run system, we made several simplifying assumptions for expediency's sake which should not be construed as desirable for a broadcast news transcription system. For example, we made no attempt to deal with telephone bandwidth data nor did we try unsupervised adaptation on the test, even though our past experience indicates these will enhance performance signicantly [4,5].…”

Section: Dry-run Systemmentioning

confidence: 99%

“…It is well known that subtracting the cepstrum mean from the input is a simple and eective procedure for removing linear dierences between channels [4]. We had unintentionally omitted this step in our dry-run system in which we adapted the seed model, estimated from WSJ acoustics, directly to the Marketplace data for each of the three prior classes.…”

Section: Adaptation To Channelmentioning

confidence: 99%

Transcribing radio news

Kubala¹,

Anastasakos²,

Jin³

et al.

Proceeding of Fourth International Conference on Spoken Language Processing. ICSLP '96

View full text Add to dashboard Cite

We h a v e recently extended the capabilities of BBN's large vocabulary discrete-utterance speech recognition system (BY-BLOS) to operate on raw audio recordings of radio news programming. The recordings are given to the system as large monolithic waveforms without any additional sideinformation. Our goal is to transcribe all speech in the input with the highest accuracy possible. The problem is very challenging because radio news programming has frequent changes in speaker, speaking style, dialect, accent, topic, channel, and environmental conditions. Furthermore, the monolithic input presents new problems for recognition algorithms and language models since all useful boundaries (such as speaker turns or sentence ends) are unknown.

show abstract

A review of large-vocabulary continuous-speech

Young

1996

IEEE Signal Process. Mag.

199

View full text Add to dashboard Cite

Adaptation to new microphones using tied-mixture normalization

Cited by 4 publications

References 7 publications

Modeling speech imitation and ecological learning of auditory-motor maps

Modeling speech imitation and ecological learning of auditory-motor maps

Transcribing radio news

A review of large-vocabulary continuous-speech

Contact Info

Product

Resources

About