Articulation Scores Research Articles

Evaluation of speech interference due to noise and reverberation

Experiments were conducted to investigate the effect of noise and reverberation on speech intelligibility and to find a simple method for evaluating the quality of speech transmission. Articulation scores and subjectively judged intelligibility of sentences were obtained in a laboratory situation together with the physical measure of the cross-correlation between the source and the listening point. The effect of broadband steady noise was examined by a monosyllable articulation test and subjective judgment of speech interference. A three-syllable articulation test was used to detect the effect of reverberation. These subjective indices showed correspondence with the source to listening point cross-correlation. The measurement of cross-correlation appeared to be effective as an objective method for estimating speech interference.

Artificial speech production device for laryngectomees controlled by jaw and tongue movements

The purpose of this study is to develop an artificial speech production device for laryngectomees. The device is composed of three parts, a measurement part for jaw and tongue movements, an estimation part for the formant frequencies, and a voice synthesizer. This paper argues that the formant frequencies of vowels can be estimated from tongue and jaw movements of the subject, and that based on this information, continuous vowels can be produced using a voice synthesizer. Curvature of midsagittal tongue surface is measured with piezoelectric film. The degree of the jaw opening is estimated by the electrical impedence between right and left buccalis. Experimental results suggest that the formant frequencies can be determined from the curvature of the tongue surface and the jaw opening. Therefore, a mapping function is introduced to obtain the formant frequencies for the synthetic speech. The amplitude and pitch of the voice are controlled manually by the person who wears the device. The articulation score of the voice produced by the device was 95% for isolated vowels and 90% for connected vowels. Results showed the usefulness of the proposed device.

LSP‐CV speech synthesis via CV‐syllable positioning rule

AbstractTraditional rule‐based synthesized speech has such problems as being unsatisfactory in articulateness and in naturalness of the rhythm inherent to the Japanese language. This paper proposes a CV‐syllable positioning rule based on the auditory evaluated data. The LSP‐CV rule speech synthesis based on the proposed principle is described. The positioning interval of CV‐syllable (i.e., the interval between the auditory perceptual timing points determined by auditory test) is affected by the succeeding syllable. Deviations are produced in the word‐top syllable, syllabic nasal, double consonant, long vowel, diphthong and devocalized sound. The deviations are represented by a correction factor to be combined into the rule in a simple form. By this rule, both rhythm and tempo can be controlled. To improve the sound articulation, the unvoiced part is excited by the original residual signal, the transient part is excited by the adaptive correlation function excitation, and the voiced part is excited by the representative correlation function excitation. As a result, a more natural rhythmicity was realized than in the case of isochronous positioning of CV‐syllable. A rule‐based synthesized speech, which is clear and easy to hear (the single sound articulation score being 95.5%), was obtained. A hardware configuration is realized in which the CV‐syllable concatenation processing is performed on μCPU and the LSP synthesis operation is performed on DSP. Then the frame period can be reduced to 3.2 ms, indicating that the interval between auditory perceptual timing points can be controlled with a high accuracy.

Very low bit rate speech coding based on joint segmentation and variable length segment quantizer

This paper presents a very low bit rate LPC vocoder based on a joint segmentation and quantization method using spectral segments having variable time length. The method exploits the nonuniform distribution of speech characteristics in the time and spectral domains. A measure of the spectral distance between a variable‐length input speech segment and a fixed‐length segment template is introduced based on linear time warping. The optimum segment boundaries and templates for a spectral sequence are efficiently determined using a dynamic programming technique so that the total spectral distortion in a voice interval is minimized. The segment templates are obtained by a sub‐optimum pattern learning method, which guarantees a monotonic decrease in distortion, using a combined segmentation and clustering technique. Experimental results for a single male speaker show that this method reduces the initial distortion by 20% and yields a sound articulation score of 78%.

騒音の語音聴力におよぼす影響 新しい５７‐Ｓ語表による

In order to obtain the basic data on effects of noise on speech hearing, the following experiment was performed. With the several combinations of sound pressure level, mixing of 57-S word list and broad band noise were examined. The results were as follows.Data of articulation score by 57 word list was same as that of 57-S word list under several dB of speech level of 57 word list.On the condition of signal vs noise ratio being 10dB in the 57-S word list, articulation score was 10% higher than that of 57 word list.

Intensive speech therapy for cleft palate children.

AbstractForty six children with deviant articulation secondary to cleft palate or other velo‐pharyngeal disorders took part in a controlled trial, to gauge the efficacy of intensive speech therapy. Subjects were randomly allocated either to attend a six‐week intensive course of speech therapy, or to continue with their conventional weekly therapy. Results indicate a significant improvement in articulation for those children receiving intensive help immediately post‐course as compared with a very slight improvement seen in the speech of the control group. This improvement achieved by children on the course was maintained throughout the two‐year follow‐up period and, although the articulation of the control group continued to improve, at the end of the two year follow‐up period the children on the course had better articulation scores.

Noise suppression signal processing using 2‐point received signals

AbstractA number of research reports has been presented on the noise suppression signal processings as one of the important technical problems. Although a reduction feeling of the noise level can be obtained by the conventional method using the one‐point received signal, it has been pointed out that the program of improving the articulation is not sufficient. to solve this problem, we propose a noise suppression signal processing method using two‐point received signals. As the first step, the noise suppression method is discussed assuming that the desired signal is uncorrelated to the noise. It is shown that the noise suppression filter, which is the optimum in the sense of least‐mean‐square error, is obtained from the correlation coefficients between the frequency components of the two‐point received signals. Then a method is shown in which the two‐point receiving noise suppression system can be constructed using FET based on the proposed principle. Finally, to verify the effectiveness of the proposed processing, the method is applied to the two‐point received speech under room noise. the articulation tests were performed for the processed and unprocessed speeches. As a result, it was seen that the syllable articulation score could be improved by 4 to 10 percent when the correlation coefficient between the two‐point received signals of the noise is 0.1 or less and that of the speech is 0.9 or more.

Aging effect on articulation score for distorted speech test

Aging effect on articulation score for distorted speech test

A nonsense syllable test designed for articulation index testing

The band‐importance function developed by Bell Laboratories is not directly applicable to any of the available recorded speech tests. For this reason we developed a test which is like that used by the Bell Laboratories but unlike other available tests in that (1) each CV or VC is a different utterance (2) the phonemes appear with uniform frequency. The same phonemes used by the Bell Laboratories were used. A “test” consists of utterances by one male and one female talker each speaking 152 CV and VC's following a carrier phrase. The frequency spectrum of this material, as well as the procedure used for its estimate, will be described. Practice effects were studied at different magnitudes of distortion. The importance function which is presumably applicable to this material was found valid within the error of measurement for the filtered speech articulation test results. The error in predicting the articulation score will be related to the magnitude of the intrinsic variability of articulation testing. [Work supported by NINCDS.]

Voice Communication System Performance in the Presence of Automotive Ignition Noise

This paper reports on recent measurements of voice system performance in a controlled automotive ignition noise environment. At a quiet measurement site in the vicinity of Tucson, AZ, various combinations of from 1 to 12 vehicles of various ages, types, and manufacture were used to provide impulsive interference to four-voice communication system types. These systems were SSB at 23 MHz, with a 10-kHz bandwidth, FM at 75 MHz with a 30-kHz bandwidth, AM at 300 MHz with a 30-kHz bandwidth, and a FM/ PCM system at 900 MHz with a 300-kHz bandwidth. The statistical characterisitcs ( amplitude probability distribution of the received noise envelope and the average crossing-rate characteristics) of the interference were measured simultaneously with the system performance. Two methods were used to measure system performance. An objective measurement using instrumentation termed voice intelligibility analysis system ( VIAS) was used which produces an intelligibility measure called articulation index ( Al). In addition, a subjective method was used employing phonetically balanced word lists and a highly trained listener panel to obtain the articulation score ( AS). One object of the measurements was to ascertain if a relationship exits between Al and AS for impulsive noise, and, indeed, quite good correlation was obtained for all four systems. The results of the system performance measurements and associated results are presented and discussed in detail in this paper.

Teaching vowel articulation with the computer vowel trainer. Methodology and results.

The Computer Vowel Trainer (CVT) uses linear prediction to estimate and display the shape of the speaker's vocal tract during an utterance. The assessment of the CVT was carried out over a period of three years with 14 profoundly deaf children divided into an experimental and a matched control group. Two main results were obtained: (a) children learning to articulate with the CVT achieve consistently higher articulation scores than those trained with conventional methods; (b) the younger the children, the greater their progress in vowel learning, retention and generalization.

Identification of synthetic speech stimuli by hearing‐impaired subjects

Identification tests of synthetic speech stimuli can provide a more direct and accurate estimate of the impairment of speech perception in hearing‐impaired subjects than those provided by the pure‐tone audiometry and discrimination tests using nonspeech stimuli [H. Fujisaki et al., Ann. Rep. Eng. Res. Inst., Fac. Eng., Univ. Tokyo 28, 74–79 (1968)]. The present study utilizes synthetic speech sounds for measuring the accuracy of identification for stationary vowels, a semivowel or a liquid, and voiced stop consonants, and compares the performances of hearing‐impaired subjects with medium‐to‐severe hearing losses to those of normal subjects. The difference in the identification accuracy between normal and hearing‐impaired subjects is found to increase as the stimuli become less stationary. While in normal subjects no significant differences are found among identification accuracy for the three kinds of stimuli, a marked deterioration is observed in hearing‐impaired subjects for less stationary stimuli. A comparison of the identification accuracy and the articulation score also reveals that the former is more sensitive than the latter as a measure of speech perception.

An audiological evaluation method of grading the speech disturbance (author's transl)

Various speech methods have been devised and applied to examine speech disturbances. However, there have been no methods to determine quantitatively the degrees of the various speech disturbances, and to decide whether they have recovered or not on the basis of the values obtained from tests. Any methods have hardly been proposed to express quantitatively conversational abilities and to evaluate objectively social adaptabilities of speech disorders.The author proposed a new quantitative method for judging the degrees of speech disturbances. The principle of this method is an auditory comparison of the speech of the patients with speech disturbances and the artificially distorted speech.Eighteen kinds (No.1…No.18) of distorted words were used in this test, which resemble speech disturbances. Standard test word sounds were classified into 6 degrees based on the auditory impression and the discrimination in pronunciation.Degree I was the normal speech (No. 1, speech articulation score 98-100%). Degree II was slightest disturbances in pronunciation (No. 2, 3 and 9). Degree III was slight disturbance (No. 7 and 8 articulation score 50-75%). Degree IV was the moderate uttering disturbances (No. 4, 10, 13, 15 and 16, articulation score 30-50%). Degree V was severe uttering disturbances (No. 5, 11, 14 and 17, articulation score 10-30%). Degree VI was the loss of speech communication (No. 6, 12 and 18, articulation score, not higher than 10%).The test was applied to 30 patients with convulsive paralytic dysarthria. The degrees of disturbances in these patients almost coincided with their speech abilities in conversation. There was no difference of scores obtained by skillful and unskilled testers. Therefore, the conversation of patients with speech disturbances could be easily classified into the grades of conversational abilities by this test.It is possible to judge easily the degrees of conversational abilities with this test when this is applied to speech disturbances in the physiological process.

A study on Prediction of Articulation Development

In an attempt to construct a short predictive test of articulation, the data obtained on 147 subjects aged 4 through 6 with misarticulation were analyzed. The subjects were those who were tested twice at an interval of two years with the comprehensive test which was specially composed for this purpose. The test consists of 509 items in 8 subtests (four articulation tests, two auditory tests, and two other tests) . The test scores of the 509 items were arranged into 45 composite scores.Correlation coefficients were computed between the articulation score (represented by word articulation test-voluntary) of the 2-year-later test, and each of 45 composite scores of the first test. Twentysix composit a scores were selected at a 1% level of significance, as ones which showed higher predictive value or correlation coefficient. Using the articulation score and 26 composite scores, a multiple correlation coefficient was computed (R=0.67), and a multiple regression equation was derived.Applying the equation to each subjest using the 26 composite scores of the first test, 147 predictive articulation scores were obtained. The predictive rate was 74.8% of observed score of the two-year-later test.

The effects of examiner race on the articulation responses of black school-aged children

Ninety 6-, 7-, and 8-year-old black children were administered the Watts Articulation Test for Screening by a white examiner and by a black examiner. The subjects' obtained significantly better articulation scores when the test was administered by the white examiner than when it was administered by the black examiner. Age of subjects and the interactions of age and sex, and of examiner age and sex were found to be significant.

Simultaneous monotic masking as a function of linguistic violations.

An investigation was conducted to determine whether grammatical, semantical, and phonological constraints affect aural masking. The rationale centered around the non-representative and non-optimal characteristics of white noise as a masker. Also, prior conclusions that content and linguistic rules do not influence masking warranted examination. Six lists, each of 25 simultaneously competing message of 36 college-age, normal hearing, native speakers of English. The 6 classes of competing messages were forward grammatical strings, forward semantically anomalous strings, forward ungrammatical strings, and the same 3 sets reproduced backwards. Articulation scores computed in percentages served as measures. Performances when the forward semantically anomalous strings and the forward ungrammatical strings functioned as maskers were depressed relative to the other conditions and not significantly different from each other.

Similarity between complexity on Role Construct Repertory Technique and articulation on Draw-A-Person test for patients and non-patients.

In an investigation of the underlying unity of the articulation of human figure drawings and personal construct complexity, moderate correlations were found, but in opposite directions, for inpatient and non-patient groups. For inpatients, greater construct complexity was associated with less sophisticated drawings. For non-patients, greater construct complexity was correlated with more sophisticated human figures. Mean and range differences in complexity and articulation scores between the two groups suggested that the higher complexity scores of the patient group indicated a loss of construct coordination, rather than greater articulation and organization of self and interpersonal experience.

Simultaneous monotic masking as a function of linguistic violations

An investigation was conducted to determine whether grammatical, semantical, and phonological constraints affect aural masking. The rationale for the study centered around conflicting reports in the literature as to the importance of content and linguistic rules for masking efficiency. Thirty-six college-age, normal-heating, native speakers of English participated in six conditions of linguistic masking. Conditions differed in type and degree of linguistic violation of the competing message. Articulation scores computed in percentages served as measures. Significant differences in performance were discussed in terms of information theory, figure-ground and closure concepts, linguistic analysis, attention, memory, and linguistic feature detection.

A STUDY ON PRESCRIPTION OF HEARING AID FOR HARD OF HEARING

The patients with sensorineural hearing loss depend greatly on the use of hearing aid for communication, because no effective treatment of perceptive deafness has been established. They usually expect much advantages by using the instruments. Therefore otologists and audiologists should request for makers to improve the hearing aid and select a suitable hearing aid for each patient. It is also necessary to do auditory training and guidance including the better use and limitations of the hearing aids.This study was performed in order to inquire the effects and the limits of hearing aid using the Master Auditory Trainer. The power and response of the modurated sound with this trainer was variable voluntarily. The hearing abilities of patients were measured by different test of speech using this trainer.The receptive articulation score of the hearing impared were increaced when they could recognize the sounds of 70-80 dB (audiometer level) filtered at 1000-2000 Hz. However, voice sounds amplified to 90 dB or more (audiometer level) did not improve significantly the score.These phenomenas related to the auditory sensation area of patients, especially to the uncomfortable hearing level.For example, in the narrow hearing area or the lower uncomfortable level, their articulation score were not improve with any amplifeid speech sound regardless their hearing loss or hearing curve.Though it is impossible to determine the frequency characteristics of the hearing aid from hearing curve in the audiogram, a favorable frequency caracteristics for the deaf can be presumed to by the average level of hearing loss. It is hoped that the prescription of the hearing aid be made. The above-mentioned facts should be taken into consideration in auditory training for the patient.

Occupational experience with a specific phraseology: Group differences in intelligibility for synthesized and human speech

Assuming a model of speech perception in which the listener's linguistic and pragmatic expectations reduce the uncertainty in the decoding process, it was hypothesized that listener‐familiarity with a particular type of phraseology would affect intelligibility of novel utterances in that phraseology. Two groups listened to sentences of two types—common phraseology and specialized phraseology appropriate for air traffice communications and cockpit annunciator messages. Subjects in the experimental group were commercial airline pilots. Those in the control group were police officers. Requirements for the two groups were equivalent regarding age, sex, education, and experience with two‐way communications using radios with poor frequency response and low S/N ratios. Half the subjects in each group heard the sentences as spoken by an air traffic controller. The other half heard the same sentences as produced by a VOTRAX speech synthesizer. Articulation scores for the common phraseology for the two occupational groups showed no significant difference. In contrast, the data showed a strong effect resulting from familiarity with the specialized phraseology. This suggests group differences in the “linguistic competence” utilized by the analysis‐by‐expectation speech perception process.