Speech database development at MIT: Timit and beyond

Zue, Victor W.; Seneff, Stephanie; Glass, James

doi:10.1016/0167-6393(90)90010-7

Cited by 508 publications

(249 citation statements)

References 1 publication

Supporting

Mentioning

236

Contrasting

Order By: Relevance

“…Four speech intervals were replaced with noise: Cs, Vs, CVs, and VCs. The TIMIT database provides phonetic classification to distinguish American English consonant and vowels (45)(46)(47); this classification was adopted with one modification. When an interval marked "closure" (labeled separately in the TIMIT database) preceded one labeled stop consonant or affricate, these intervals were combined and treated as a single consonant.…”

Section: Methodsmentioning

confidence: 99%

Cochlea-scaled entropy, not consonants, vowels, or time, best predicts speech intelligibility

Stilp

Kluender²

2010

Proc. Natl. Acad. Sci. U.S.A.

100

View full text Add to dashboard Cite

Speech sounds are traditionally divided into consonants and vowels. When only vowels or only consonants are replaced by noise, listeners are more accurate understanding sentences in which consonants are replaced but vowels remain. From such data, vowels have been suggested to be more important for understanding sentences; however, such conclusions are mitigated by the fact that replaced consonant segments were roughly one-third shorter than vowels. We report two experiments that demonstrate listener performance to be better predicted by simple psychoacoustic measures of cochlea-scaled spectral change across time. First, listeners identified sentences in which portions of consonants (C), vowels (V), CV transitions, or VC transitions were replaced by noise. Relative intelligibility was not well accounted for on the basis of Cs, Vs, or their transitions. In a second experiment, distinctions between Cs and Vs were abandoned. Instead, portions of sentences were replaced on the basis of cochlea-scaled spectral entropy (CSE). Sentence segments having relatively high, medium, or low entropy were replaced with noise. Intelligibility decreased linearly as the amount of replaced CSE increased. Duration of signal replaced and proportion of consonants/vowels replaced fail to account for listener data. CSE corresponds closely with the linguistic construct of sonority (or vowel-likeness) that is useful for describing phonological systematicity, especially syllable composition. Results challenge traditional distinctions between consonants and vowels. Speech intelligibility is better predicted by nonlinguistic sensory measures of uncertainty (potential information) than by orthodox physical acoustic measures or linguistic constructs.

show abstract

Section: Methodsmentioning

confidence: 99%

Cochlea-scaled entropy, not consonants, vowels, or time, best predicts speech intelligibility

Stilp

Kluender²

2010

Proc. Natl. Acad. Sci. U.S.A.

100

View full text Add to dashboard Cite

show abstract

“…The single word stimuli in the database include repetitions of English digits, the international radio alphabets, the 20 most frequent words in the British National Corpus (BNC), and a set of words selected by Kent et al to demonstrate relevant phonetic contrasts [9]. The sentence stimuli are derived from the Yorkston-Beukelman assessment of intelligibility [10] and the TIMIT database [11]. In addition, each participant is asked to describe the contents of a few photographs that are selected from standardized tests of linguistic ability in his/her own words so as to include dictationstyle speech into the database.…”

Section: Description Of Datamentioning

confidence: 99%

Adapting acoustic and lexical models to dysarthric speech

Mengistu

Rudzicz

2011

2011 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

Dysarthria is a motor speech disorder resulting from neurological damage to the part of the brain that controls the physical production of speech and is, in part, characterized by pronunciation errors that include deletions, substitutions, insertions, and distortions of phonemes. These errors follow consistent intra-speaker patterns that we exploit through acoustic and lexical model adaptation to improve automatic speech recognition (ASR) on dysarthric speech. We show that acoustic model adaptation yields an average relative word error rate (WER) reduction of 36.99% and that pronunciation lexicon adaptation (PLA) further reduces the WER by an average of 8.29% relative on a large vocabulary task of over 1500 words for 6 speakers with severe to moderate dysarthria. PLA also shows an average relative WER reduction of 7.11% on speaker-dependent models evaluated using 5-fold crossvalidation.

show abstract

“…The only existing corpus of regional variation in the United States that obtained high quality audio recordings in a sound-attenuated booth is the TIMIT Acoustic-Phonetic Continuous Speech Corpus (Fisher et al, 1986;Zue et al, 1990). The TIMIT corpus contains recordings of 630 talkers who each read 10 different sentences.…”

Section: Existing Spoken Language Corpora With Dialect Variationmentioning

confidence: 99%

The Nationwide Speech Project: A new corpus of American English dialects

Clopper

Pisoni

2006

Speech Communication

View full text Add to dashboard Cite

Perceptual and acoustic research on dialect variation in the United States requires an appropriate corpus of spoken language materials. Existing speech corpora that include dialect variation are limited by poor recording quality, small numbers of talkers, and/or small samples of speech from each talker. The Nationwide Speech Project corpus was designed to contain a large amount of speech produced by male and female talkers representing the primary regional varieties of American English. Five male and five female talkers from each of six dialect regions in the United States were recorded reading words, sentences, passages, and in interviews with an experimenter, using high quality digital recording equipment in a sound-attenuated booth. The resulting corpus contains nearly an hour of speech from each of the 60 talkers that can be used in future research on the perception and production of dialect variation.

show abstract

Speech database development at MIT: Timit and beyond

Cited by 508 publications

References 1 publication

Cochlea-scaled entropy, not consonants, vowels, or time, best predicts speech intelligibility

Cochlea-scaled entropy, not consonants, vowels, or time, best predicts speech intelligibility

Adapting acoustic and lexical models to dysarthric speech

The Nationwide Speech Project: A new corpus of American English dialects

Contact Info

Product

Resources

About