A Bayesian model of continuous speech recognition is presented. It is based on Shortlist (D. Norris, 1994; D. Norris, J. M. McQueen, A. Cutler, & S. Butterfield, 1997) and shares many of its key assumptions: parallel competitive evaluation of multiple lexical hypotheses, phonologically abstract prelexical and lexical representations, a feedforward architecture with no online feedback, and a lexical segmentation algorithm based on the viability of chunks of the input as possible words. Shortlist B is radically different from its predecessor in two respects. First, whereas Shortlist was a connectionist model based on interactive-activation principles, Shortlist B is based on Bayesian principles. Second, the input to Shortlist B is no longer a sequence of discrete phonemes; it is a sequence of multiple phoneme probabilities over 3 time slices per segment, derived from the performance of listeners in a large-scale gating study. Simulations are presented showing that the model can account for key findings: data on the segmentation of continuous speech, word frequency effects, the effects of mispronunciations on word recognition, and evidence on lexical involvement in phonemic decision making. The success of Shortlist B suggests that listeners make optimal Bayesian decisions during spoken-word recognition.

/pdf/shortlist-b-a-bayesian-model-of-continuous-speech-zxk04ogtk2.pdf

Shortlist B: A Bayesian model of continuous speech recognition

This article presents a review of the effects of adverse conditions (ACs) on the perceptual, linguistic, cognitive, and neurophysiological mechanisms underlying speech recognition. The review starts with a classification of ACs based on their origin: Degradation at the source (production of a noncanonical signal), degradation during signal transmission (interfering signal or medium-induced impoverishment of the target signal), receiver limitations (peripheral, linguistic, cognitive). This is followed by a parallel, yet orthogonal classification of ACs based on the locus of their effect: Perceptual processes, mental representations, attention, and memory functions. We then review the added value that ACs provide for theories of speech recognition, with a focus on fundamental themes in psycholinguistics: Content and format of lexical representations, time-course of lexical access, word segmentation, feed-back in speech perception and recognition, lexicalsemantic integration, interface between the speech system and general cognition, neuroanatomical organisation of speech processing. We conclude by advocating an approach to speech recognition that includes rather than neutralises complex listening environments and individual differences.

/pdf/speech-recognition-in-adverse-conditions-a-review-1gdpt01sgx.pdf

Speech recognition in adverse conditions: A review

Lexical competition in non-native spoken-word recognition

Successful speech perception requires that listeners map the acoustic signal to linguistic categories. These mappings are not only probabilistic, but change depending on the situation. For example, one talker's /p/ might be physically indistinguishable from another talker's /b/ (cf. lack of invariance). We characterize the computational problem posed by such a subjectively nonstationary world and propose that the speech perception system overcomes this challenge by (a) recognizing previously encountered situations, (b) generalizing to other situations based on previous similar experience, and (c) adapting to novel situations. We formalize this proposal in the ideal adapter framework: (a) to (c) can be understood as inference under uncertainty about the appropriate generative model for the current talker, thereby facilitating robust speech perception despite the lack of invariance. We focus on 2 critical aspects of the ideal adapter. First, in situations that clearly deviate from previous experience, listeners need to adapt. We develop a distributional (belief-updating) learning model of incremental adaptation. The model provides a good fit against known and novel phonetic adaptation data, including perceptual recalibration and selective adaptation. Second, robust speech recognition requires that listeners learn to represent the structured component of cross-situation variability in the speech signal. We discuss how these 2 aspects of the ideal adapter provide a unifying explanation for adaptation, talker-specificity, and generalization across talkers and groups of talkers (e.g., accents and dialects). The ideal adapter provides a guiding framework for future investigations into speech perception and adaptation, and more broadly language comprehension.

Robust speech perception: recognize the familiar, generalize to the similar, and adapt to the novel.

This paper presents the case of a 17-year-old right-handed Belgian boy with developmental FAS and comorbid developmental apraxia of speech (DAS). Extensive neuropsychological and neurolinguistic investigations demonstrated a normal IQ but impaired planning (visuo-constructional dyspraxia). A Tc-99m-ECD SPECT revealed a significant hypoperfusion in the prefrontal and medial frontal regions, as well as in the lateral temporal regions. Hypoperfusion in the right cerebellum almost reached significance. It is hypothesized that these clinical findings support the view that FAS and DAS are related phenomena following impairment of the cerebro-cerebellar network.

/pdf/developmental-foreign-accent-syndrome-report-of-a-new-case-2yllsms4ct.pdf

Developmental Foreign Accent Syndrome : Report of a new case

An evaluation of vowel normalization procedures for the purpose of studying language variation is presented. The procedures were compared on how effectively they (a) preserve phonemic information, (b) preserve information about the talker's regional background (or sociolinguistic information), and (c) minimize anatomical/physiological variation in acoustic representations of vowels. Recordings were made for 80 female talkers and 80 male talkers of Dutch. These talkers were stratified according to their gender and regional background. The normalization procedures were applied to measurements of the fundamental frequency and the first three formant frequencies for a large set of vowel tokens. The normalization procedures were evaluated through statistical pattern analysis. The results show that normalization procedures that use information across multiple vowels ("vowel-extrinsic" information) to normalize a single vowel token performed better than those that include only information contained in the vowel token itself ("vowel-intrinsic" information). Furthermore, the results show that normalization procedures that operate on individual formants performed better than those that use information across multiple formants (e.g., "formant-extrinsic" F2-F1).

/pdf/a-comparison-of-vowel-normalization-procedures-for-language-4l7semooai.pdf

A comparison of vowel normalization procedures for language variation research.

Native American English and non-native (Dutch) listeners identified either the consonant or the vowel in all possible American English CV and VC syllables. The syllables were embedded in multispeaker babble at three signal-to-noise ratios (0, 8, and 16 dB). The phoneme identification performance of the non-native listeners was less accurate than that of the native listeners. All listeners were adversely affected by noise. With these isolated syllables, initial segments were harder to identify than final segments. Crucially, the effects of language background and noise did not interact; the performance asymmetry between the native and non-native groups was not significantly different across signal-to-noise ratios. It is concluded that the frequently reported disproportionate difficulty of non-native listening under disadvantageous conditions is not due to a disproportionate increase in phoneme misidentifications.

/pdf/patterns-of-english-phoneme-confusions-by-native-and-non-50wp1ch4co.pdf

Patterns of English phoneme confusions by native and non-native listeners

A database is presented of measurements of the fundamental frequency, the frequencies of the first three formants, and the duration of the 15 vowels of Standard Dutch as spoken in the Netherlands (Northern Standard Dutch) and in Belgium (Southern Standard Dutch). The speech material consisted of read monosyllabic utterances in a neutral consonantal context (i.e., /sVs/). Recordings were made for 20 female talkers and 20 male talkers, who were stratified for the factors age, gender, and region. Of the 40 talkers, 20 spoke Northern Standard Dutch and 20 spoke Southern Standard Dutch. The results indicated that the nine monophthongal Dutch vowels /a ɑ ɛ i ɪ ɔ u y ʏ/ can be separated fairly well given their steady-state characteristics, while the long mid vowels /e o o/ and three diphthongal vowels /ɛɪ ɔu y/ also require information about their dynamic characteristics. The analysis of the formant values indicated that Northern Standard Dutch and Southern Standard Dutch differ little in the formant frequencies at steady-state for the nine monophthongal vowels. Larger differences between these two language varieties were found for the dynamic specifications of the three long mid vowels, and, to a lesser extent, of the three diphthongal vowels.

/pdf/an-acoustic-description-of-the-vowels-of-northern-and-4175r7tmts.pdf

An acoustic description of the vowels of Northern and Southern Standard Dutch.

https://pure.mpg.de/pubman/item/item_60320_3/component/file_60321/vanAlphen_2004_acoustical.pdf

Roel Smits

Papers

A comparison of vowel normalization procedures for language variation research.

Patterns of English phoneme confusions by native and non-native listeners

An acoustic description of the vowels of Northern and Southern Standard Dutch.

Acoustical and perceptual analysis of the voicing distinction in Dutch initial plosives: The role of prevoicing

Supervised and unsupervised learning of multidimensionally varying non-native speech categories