It is well known that listeners can ignore disturbances in speech and rely on context to interpolate the message.This fact is used to determine the importance of individual words for projecting Transition Relevance Places, TRPs.Subjects were asked to shadow manipulated pre-recorded dialogs with minimal responses, saying 'ah' when they feel it is appropriate.In these dialogs, at random, of each utterance, either one of the last four words was replaced by white noise (masked condition), or no word was replaced (non masked condition).The reaction times were analyzed for effects of masked words.The presence of masked words, even prominent words, did not affect the response times of our subjects unless the very last word of the utterance was masked.This indicates that listeners are able to seamlessly interpolate the missing words and only need the identity of the last word to determine the exact position of the TRP.

Recent studies of spoken Standard Dutch support an ongoing change in the phonetic quality of the diphthong /EI/ [1,2].However, there is a need for broader analyses and larger data sets.Here, we took Dutch vowel variants of 44 speakers from a spoken Dutch speech corpus, the CGN [3].The vowels were measured and compared on the basis of 15.000 vowel segments, consisting of productions of /EI/, /Au/, /2y/, /o:/, and /e:/, as well as the anchor vowels /a/, /i/, /u/.It was our aim to analyze changes in vowel quality dependent on the speakers' sociological backgrounds and ages, and to deal with the variable recording qualities of the corpus.All vowels were taken from spontaneously uttered sentences and were analyzed automatically by means of a principal component analysis (PCA) on the vowels' bark-filtered spectra, as well as by formant analysis.Recalculating spectral positions in the principal components (pc's) plane displayed the spectral interaction of the first formants in the pc1-pc2 plane, and explained the better separability of the vowels compared to the F1-F2 plane, as well as the high correlation of the first three formants with pc1 and pc2.The first pc's turned out to be rather insensitive to sex-differences, but they were sensitive to the signal-to-noise ratio of the speech data.Variable recording qualities manifested themselves in speaker-specific location and size of the vowel spaces.Good signal-to-noise ratios could be transformed to poorer signals by increasing the lowest possible dB values per filter.Having analyzed the influence of noise on our data, we could normalize the data by taking each speaker's /a-i-u/ positions and the focal point as references for better inter-speaker comparison.The results clearly show different vowel quality patterns dependent on the speakers' education and age, and indicate a progress of quality changes with as parameters the lowering and the degree of diphthongization of the long vowels and diphthongs.

The effect of the position of the last accented word on the projection of TRPs was investigated with two RT experiments.Subjects were asked to respond with minimal responses to prerecorded dialogs and impoverished versions of these dialogs, containing either only intonation and pause information,hummed stimuli, or no periodic component at all, whispered stimuli.The distribution of these elicited response delays was comparable to that of natural turn switches.It is shown that the presence of non-prominent words before a TRP reduces the delays of elicited and natural responses alike, even in impoverished speech.This suggests that the presence of an prominent, informative, word starts the projection of a possible upcoming TRP.The availability of non-prominent, predictable, speech then allows listeners to improve their predictions of the exact timing of the TRP.

In two RT experiments, subjects were asked to respond with minimal responses to prerecorded dialogs and impoverished versions of these dialogs, containing either only intonation and pause information, hummed stimuli, or no periodic component at all, whispered stimuli. For the hummed, stimuli, response delays and, especially, variances were higher than the original recordings. Responses to mid-frequency pitch utterance-ends were significantly longer than responses to low pitch utterance-ends, suggesting that our subjects fell back to reacting to pauses when presented with hummed utterances ending in a mid-frequency tone. This suggests that, in contrast to low or high end-tones, intonation contours that end in a mid-frequency tone might not contain any useful information for predicting end-of-utterance TRPs. We conclude that just the intonation and pauses of a conversation contain sufficient information for projection of TRPs. However this information is measurably impoverished with respect to original to an extent that increases the “processing ” time by 10%. No difference was found between whispered and original speech. This lack of any effect of removing all periodic sound components from the speech signal indicates that in natural speech the pitch signal itself might be redundant for predicting TRPs. 1.

This paper is an initial report on the systematic analysis of changes within the vowel system of Standard Dutch. The work focuses on the recent lowering of the diphthong /Ei/, known as ‘Polder Dutch ’ (Poldernederlands). The purpose was to find an automatizable method to reliably analyze and compare speakers of a large corpus of Dutch spon-taneous speech. Diphthong variants of twelve native speakers were compared by measuring both formants and spectral en-ergy distributions. The resulting pc1-pc2 plane of a principal component analysis (PCA) on the barkfiltered spectrum of an-chor vowels /a/, /i/, /u / was comparable to the F1-F2 bark plane of the same segments and explained 90 % of the variance in the data. The differences between the speakers ’ vowel spaces ap-peared not to depend on sex, so male and female data could be calculated within the same analysis. The distance of the diphthong /Ei / to the anchor vowels in the pc1-pc2 plane, together with the dipthong dynamics were the major cues to diphthong quality. Further indication of a sound change in progress for the ‘Polder Dutch ’ speakers was the po-sition of /E / in the vowel space. 1.

A total laryngectomy changes the anatomy and physiology of the vocal tract, with a most noticeable effect on speech.By applying a voice prosthesis, enabling the patient to use tracheoesophageal (TE) speech, speech is of better quality than with esophageal or electrolarynx speech, but still very deviant from laryngeal speech.Most studies on TE speech have focused on voice quality.For the understanding of the physiology of the neoglottis and for improving the results of rehabilitation, it is important to study intelligibility as well.This paper will discuss the first results of a study on TE speech intelligibility.

Traditional hand-edited formant measurements may result in biased assessment of vowel formants in children’s speech. Therefore, vowel spaces that are constructed by hand-edited formant measures may be unreliable. The recent development of an automated frequency domain analysis method allows for more reliable measurements. Thus, a valid comparison of the size and positioning of young children’s vowel spaces across languages can be achieved. Contrasting the extension of the vowel space utilized by young children acquiring Hungarian and Dutch can provide information pertaining to a) children’s abilities to explore the vowel space and b) potential crosslinguistic differences in exploiting the potentially available acoustic vowel space. Since a unified theory of vowel acquisition has never been developed, it is hoped that the new method will contribute to the creation of such a theory by comparing and contrasting results from diverse languages. Results suggest that two-year-old Hungarian- and Dutchspeaking children utilize the vowel space languagespecifically, by exploiting different regions within the potentially available acoustic space.

Speech is considered an efficient communication channel. This implies that the organization of utterances is such that more speaking effort is directed towards important parts than towards redundant parts. Based on a model of incremental word recognition, the importance of a segment is defined as its contribution to word-disambiguation. This importance is measured as the segmental information content, in bits. On a labeled Dutch speech corpus it is then shown that crucial aspects of the information structure of utterances partition thesegmental information content and explain 90% of the variance. Two measures of acoustical reduction, duration and spectral center of gravity, are correlated with the segmental information content in such a way that more important phonemes are less reduced. It is concluded that the organization of conventional information structure does indeed increase efficiency.

Speaking is generally considered efficient in that less effort is spent articulating more redundant items.With efficient speech production, less reduction is expected in the pronunciation of phonemes that are more important (distinctive) for word identification.The importance of a single phoneme in word recognition can be quantified as the information (in bits) it adds to the preceding word onset to narrow down the lexical search.In our study, segmental information showed to correlate consistently with two measures of reduction: vowel duration and formant reduction.This correlation was found after accounting for speaker and vowel identity, speaking style, lexical stress, modeled prominence, and position of the syllable in the word.However, consistent correlations are only found in high-frequency words.Furthermore, the correlation is strongest in normal reading and weaker in spontaneous and anomalous read speech.Combined, these facts suggest that this type of efficiency in production might rely on retrieving stored words from memory.Efficiency in vowel production seems to be less or absent when words have to be assembled on-line.

This paper summarizes the results of a series of experiments conducted to investigate various aspects of normal pharyngeal articulation and the nature of pharyngeal coarticualtion.Video fiberscopic imaging, electromagnetography and acoustic analysis techniques were used to obtain empirical and quantitative data on the use of the pharynx in speech production.The overall results suggest that mechanical constraints determine to a great extent the articulatory strategy used by the speaker to achieve the perceptual/acoustic contrast essential for the process of speech encoding.

Copyright It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content licence (like

An open source database of hand-segmented Dutch speech was constructed with off-the-shelf software using speech from 8 speakers in a variety of speaking styles.For a total of 50,000 words, speech acquisition and preparation took around 3 person-weeks per speaker.Hand segmentation took 1,000 hours of labeling altogether.The asymptotic segmentation speed was about one word, or four boundaries, per minute.An evaluation showed that the Median Absolute Difference of the segment boundaries was 6 ms between labelers, and 4 ms within labelers.Label differences (substitutions, insertions, and deletions) were found in 8% of the segments between labelers and 5% within labelers.Compiled data are available in relational database format for querying with SQL.

In this study acoustical as well as lexical/syntactic correlates of prominence are analyzed and discussed.Prominence is defined at the word level and is based on listener judgments.Spoken sentences from many different speakers, taken from the Dutch Polyphone corpus of telephone speech, are analyzed.A selection of useful acoustical input features is chosen for classification of word prominence, by means of Feed Forward Nets.For an independent test set of 1,000 sentences about 79% of the words are correctly classified as prominent or not.We also developed an algorithm, based on text input, using linguistic/syntactical features derived from text only, to predict prominence.The prediction agrees with the perceived prominence in 81% of the cases for the independent test set.The results of this project show that certain acoustical and linguistic correlates of prominence can be extracted automatically and can be used to accurately predict prominence with a consistency similar to the pominence assignment by naive listeners.

A large part of the variation in natural speech appears along the dimensions of articulatory precision / perceptual distinctiveness. We propose that this variation is the result of an effort to communicate efficiently. Speaking i s considered efficient if the speech sound contains only the information needed to understand it. This efficiency is tested by means of a corpus of spontaneous and matched read speech, and syllable and word frequencies as measures of information content (12007 syllables, 8046 word forms, 1582 intervocalic consonants, and 2540 vowels). It is indeed found that the duration and spectral reduction of consonants and vowels correlate with the frequency of syllables and words in this corpus. Consonant intelligibility correlates with both the acoustic factors and the syllable and word frequencies. It i s concluded that the principle of efficient communication organizes at least some aspects of speech production.

This paper describes a first step towards the automatic classification of prominence (as defined by native listeners). As a result of a listening experiment each word in 500 sentences was marked with a rating scale between `0' (non-prominent) and `10' (very prominent). These prominence labels are compared with the following acoustical features: loudness of each vowel, and F0 range and duration of each syllable. A linear relationship between the rating scale of prominence and these acoustical features is found. These acoustical features then are used for a preliminary automatic classification to predict prominence.

Reduction causes changes in the acoustics of consonant realizations that affect their identification. In this study we try to identify some of the acoustic parameters that are correlated with this change in identification. Speaking style is used to manipulate the degree of reduction. Pairs of otherwise identical intervocalic consonants from read and spontaneous utterances are presented to subjects in an identification experiment. The resulting identification scores are correlated to five different acoustical measures that are affected by the amount of consonant reduction: Segmental duration, spectral Center of Gravity, intervocalic sound energy difference, intervocalic F 2 slope difference, and the amount of vowel reduction in the syllable kernel. The identification differences between the read and spontaneous realizations are compared with the differences in each of the acoustic measures. It showed that only segmental duration and the spectral Center of Gravity are significantly corre...

acoustic signal which simply represents the sequence of the actually pronounced phones. This might in turn reduce recognition accuracy. One solution to this problem is to use ([2], for both French and English, and [3]) which account for various pronunciation variations in the language. However, it is hard to obtain reliable sets of such rules. Because of the compactness of such rules, many variation phenomena cannot be covered. In this study, we develop data-based word juncture models, which account for the pronunciation variations at word boundaries, as an optional form of phonological rules. We used the American English TIMIT database. Issues in generating the models and using them in a continuous recognition task are discussed. A comparison is given between the coverage of the pronunciation variations by the models and by a set of phonological rules. There is a fairly good agreement between the models and the rules in predicting the pronunciation variations, whereas the models cover a larger set of variation phenomena. Furthermore, use of the models improved recognition performance. In this study, we propose a different approach. We only concentrate on the pronunciation variations at word junctures, although generally within-word deviations may also occur. We directly generate word-juncture models based on the statistics of the pronunciation variations from a training data set. Then we use these models to predict those actual phone sequences deviating from the normative sequences. The actual phone sequences are used in word recognition. The whole process of generating and using the models is illustrated in Figure 1.

HMMThis paper presents research on integrating context-dependent durational knowledge into HMM-based speech recognition.The first part of the paper presents work on obtaining relations between the parameters of the context-free HMMs and their durational behaviour, in preparation for the context-dependent durational modelling presented in the second part.Duration integration is realised via rescoring in the post-processing step of our N-best monophone recogniser.We use the multi-speaker TIMIT database for our analyses.The single-state dpdf (which is geometrical) is less important than the dpdf of the whole HMM, because in actual practice it is the latter that models a phonetic segment.In this section we firstly derive the closed-form whole-model dpdf for general leftto-right HMM.Left-to-right is by far the most common type of transition topology used for speech recognition.It may include any number of skipping transitions and parallel paths but no feedback loops that contains more than one state.Then an analysis will be given of the general properties of the dpdf, with the help of some examples of useful topologies.

In a joint project called SPIN, which is sponsored by the European Information Technology Program ESPRIT, a speech interface for office automation will be developed and tested. Two specific aspects of such an interface, which are discussed in this paper, have to do with speech store-and-forward and text-to-speech synthesis-by-rule. We will diagnostically evaluate the speech quality of the medium-band coders and of the diphone-based synthesis systems in their various developmental phases from software simulation to hardware prototype. This includes intelligibility, subjective quality, as well as speaker identifiability. Since only native listeners can be used to evaluate the synthetic speech quality of a text-to-speech system in a specific language, we developed a portable, stand-alone, multi-subject system which allows efficient testing on the spot.

Contrary to human speech perception, most speech-analysis and -processing techniques for front-end automatic speech recognition are based on discrete spectral analysis without implying temporal continuity, dynamic aspects, or context dependency. Speech recognition is presently primarily based on one specific signal aspect at a time, whereas one should make more efficient use of parallel information channels handling multiple cues. The actual global and local conditions should specify which single or combined set of parameters is important at that moment. Dynamic information is worth to receive more attention. The human ear actually seems to be 'overpowered' to process good quality speech; only under critical conditions the full analyzing capability of the ear and all redundant information in the speech signal have to be used. Automatic speech recognition systems not yet have this 'overcapacity' and therefore should rely more on multiple, highly resistant cues.

Participating members of the international NATO Research Study Group RSG-10 on Speech Processing are presently using a data base of connected digits, spoken in different languages, to facilitate comparison of (connected) word recognition systems in the various countries. In order to be able to refer "system" results to human performance, we executed a listening experiment with a representative subset of the same recordings of connected digits. Four Dutch subjects listened to connected 3-to-5 digit groups, as well as to isolated digits, spoken in English and in Dutch. The English material was spoken by 4 native and 6 nonnative speakers of English. Apart from an undisturbed condition, subjects also identified the digit sequences in two noise conditions with speech-to-noise ratios of -3 and -9 dB. At SNR = -3 dB the listeners still do an excellent job. There is substantial speaker variation, but no systematic effect of language, sex, or native vs non-native speakers. Subjects showed a prolonged learning effect, and were especially sensitive to tempo under more difficult (noisy) listening conditions.

The intelligibility of speech, in terms of sentence intelligibility, at low levels of noise is almost perfect, although some redundancy may already have been lost, which could result in a reduced listening comfort. In order to study this in conditions representative of normal communication, we used the intelligibility of consonants in vowel-consonant-vowel segments excerpted from conversational speech as a sensitive measure. The level of the noise while listening was systematically varied. The consonant intelligibility score, averaged over consonants (18), speakers (5), and listeners (10), was about 70% correct in a quiet condition, and started to degrade at a signal-to-noise ratio of 10 to 7 dB. Results furthermore show that when the speaker himself is also in the noise he does not just talk louder but also more intelligible. However, providing the listener with a visual image of the speaker does not improve intelligibility.

First a survey is given of a number of published vowel and word recognition systems. Then a new real-time word recognition system is described that uses only a small computer (8K memory) and a few analog peripherals. The essentials of the procedure are as follows. During the pronunciation of a word, a spectral analysis is carried out by a bank of 17 1/3-octave bandpass filters. The outputs of the filters are logarithmically amplified and the maximal amplitude of the envelope is determined and sampled every 15 ms. In this way a word is characterized by a sequence of sample points in a 17-dimensional space. Then a principal components analysis is performed, reducing the original 17 dimensions of the space to 3. After a linear time normalization, the 3-dimensional trace of the spoken word is compared with 20 reference traces, representing the 20 possible utterances (the digits, plus 10 computer commands). The machine responds by naming the best fitting trace. With the 20 speakers of the design set, the machine is correct 98.8 percent of the time.