Evaluation Of Speech Intelligibility Research Articles

A metric for evaluating speech intelligibility and quality in hearing aids

Noise and distortion reduce speech intelligibility and quality in hearing aids, but there are no metrics that encompass both noise and distortion. This presentation introduces new intelligibility and sound-quality calculation procedures based on the Speech Intelligibility Index [ANSI S3.5-1997]. The SII involves measuring the signal-to-noise ratio (SNR) in separate frequency bands, modifying the estimated noise levels to include auditory masking, and computing a weighted sum across frequency of the modified SNR values. In the new procedure, the estimated SNR is replaced by a signal-to-distortion ratio (SDR) computed from the coherence between the input and output signals of the system under test. The SDR replaces the system noise estimate with the combined noise and nonlinear distortion. For the procedure, the signal is divided into three regions comprising the low-, mid-, and high-level signal segments. The SII is then calculated for each region using the corresponding coherence. The three coherence SII values are then combined to predict the intelligibility and the sound quality for the device under test. The coherence SII procedure is shown to be accurate for both normal-hearing and hearing-impaired listeners for additive noise, peak-clipping distortion, and center-clipping distortion. [Work supported by GN ReSound (JMK) and the Whitaker Foundation (KHA).]

Method for Evaluating the Clarity of Speech Received Through the Channel of Intercommunication Devices of Self-Contained Respirators

A method is described for evaluating the clarity of speech received through the intercommunication channels of self-contained respirators based on transforming the speech communication into an electrical signal, then into digital form and use of a PC for formant evaluation of speech clarity and intelligibility.

Full Length Article

The purpose of this study was to evaluate speech intelligibility after a dehiscence operation using artificial grafts for patients with speech disorders after partial glossectomy that were caused by scars resulting from the primary operation. The subjects were six men and three women, who had had a partial glossectomy for tongue cancer followed by direct closure without reconstruction. They were operated on a second time operation to mobilise the residual tongue by dividing the cicatrix. An artificial graft was applied to the wound to maintain the dehiscence. Speech intelligibility was evaluated by a standardised Japanese speech intelligibility test before, and 6 and 12 months after the second operation. The intelligibility scores significantly improved during the first 6 months after the second operation, and continued to improve slightly during the following 6 months. This study suggests that the dehiscence operation using an artificial graft could improve speech in patients after partial glossectomy.

Evaluation of speech intelligibility after a secondary dehiscence operation using an artificial graft in patients with speech disorders after partial glossectomy

The purpose of this study was to evaluate speech intelligibility after a dehiscence operation using artificial grafts for patients with speech disorders after partial glossectomy that were caused by scars resulting from the primary operation. The subjects were six men and three women, who had had a partial glossectomy for tongue cancer followed by direct closure without reconstruction. They were operated on a second time operation to mobilise the residual tongue by dividing the cicatrix. An artificial graft was applied to the wound to maintain the dehiscence. Speech intelligibility was evaluated by a standardised Japanese speech intelligibility test before, and 6 and 12 months after the second operation. The intelligibility scores significantly improved during the first 6 months after the second operation, and continued to improve slightly during the following 6 months. This study suggests that the dehiscence operation using an artificial graft could improve speech in patients after partial glossectomy.

The effect of flattening fundamental frequency (FO) on the speech recognition and intelligibility evaluation of Mandarin

The effect of flattening fundamental frequency (FO) on the speech recognition and intelligibility evaluation of Mandarin

“I can hear”: a system for universal hearing screening in school age children. Organization and first results

Hearing screening programs in school age children are necessary for early detection and treatment of hearing impairments in this population. The Institute of Physiology and Pathology of Hearing have developed and implemented hearing screening program at schools in Poland based on multimedia hearing screening system “I can hear”. The system consists of a computer program installed on PC, an acoustic calibrator and earphones. The testing procedure involves automated questionnaire analysis, pure tone audiometric test and the evaluation of speech intelligibility in noise. The tests are conducted in schools under a supervision of specially trained screeners. In the years 1999–2001, 125,248 children were screened in Poland. A total of 71,940 results from selected regions were statistically analyzed. Eighteen percent of them did not pass the screening test. Only 29% of the children referred for diagnostic evaluation appeared at audiological centers. In one third of them, hearing impairment was confirmed.

Correlation of subjective and objective measures of speech intelligibility

Currently there are a number of objective evaluation methods used to quantify the speech intelligibility in a built environment, including the Speech Transmission Index (STI), Rapid Speech Transmission Index (RASTI), Articulation Index (AI), and the Percentage Articulation Loss of Consonants (%ALcons). Many of these have been used for years; however, questions remain about their accuracy in predicting the acoustics of a space. Current widely used software programs can quickly evaluate STI, RASTI, and %ALcons from a measured impulse response. This project compares subjective human performance on modified rhyme and phonetically balanced word tests with objective results calculated from impulse response measurements in four different spaces. The results of these tests aid in understanding performance of various methods of speech intelligibility evaluation. [Work supported by the Univ. of Nebraska Center for Building Integration.] For Speech Communication Best Student Paper Award.

NEW PRODUCT DEVELOPMENTS

Hearing screening programs in school age children are necessary for early detection and treatment of hearing impairments in this population. The Institute of Physiology and Pathology of Hearing have developed and implemented hearing screening program at schools in Poland based on multimedia hearing screening system “I can hear”. The system consists of a computer program installed on PC, an acoustic calibrator and earphones. The testing procedure involves automated questionnaire analysis, pure tone audiometric test and the evaluation of speech intelligibility in noise. The tests are conducted in schools under a supervision of specially trained screeners. In the years 1999–2001, 125,248 children were screened in Poland. A total of 71,940 results from selected regions were statistically analyzed. Eighteen percent of them did not pass the screening test. Only 29% of the children referred for diagnostic evaluation appeared at audiological centers. In one third of them, hearing impairment was confirmed.

Diagnostic articulation tables

In recent years, considerable progress has been made in the development of instrumental methods for general speech quality and intelligibility evaluation on the basis of modeling the auditory perception of speech and measuring the signal-to-noise ratio. Despite certain advantages (fast measurement procedures with a low labor consumption), these methods are not universal and, in essence, secondary, because they rely on the calibration based on subjective-statistical measurements. At the same time, some specific problems of speech quality evaluation, such as the diagnostics of the factors responsible for the deviation of the speech quality from standard (e.g., accent features of a speaker or individual voice distortions), can be solved by psycholinguistic methods. This paper considers different kinds of diagnostic articulation tables: tables of minimal pairs of monosyllabic words (DRT) based on the Jacobson differential features, tables consisting of multisyllabic quartets of Russian words (the choice method), and tables of incomplete monosyllables of the _VC/CV_ type (the supplementary note method). Comparative estimates of the tables are presented along with the recommendations concerning their application.

Speech Discrimination in Background Noise with the Vibrant® Soundbridge™ Middle Ear Implant

Objective: Evaluate speech intelligibility in competing background noise for monaural users of the Vibrant^® Soundbridge™ (VSB) middle ear implant (Symphonix). Patients and Material: Nine adults with moderate to severe bilateral sensorineural hearing loss, and implanted with a unilateral VSB, were included in the study. The patients’ experience of the implant ranged from 9 to 24 months. Retrospective single-subject repeated measures of speech intelligibility in background noise were performed using 2 different test procedures, the Elbaz test and the Garin and Galle test. The patients were tested VSB aided and VSB unaided in a variety of signal-to-noise ratios. The correct-score percentage for each ratio for each condition was averaged for the group of patients and compared with normative data for each test method. Results: The VSB-aided condition showed results approaching normative data for both test methods. At a signal-to-noise ratio of zero, the Elbaz test showed a mean correct score of 55% for the VSB-unaided condition versus 85% for the VSB-aided condition. The Garin and Galle test showed a mean score of 34% for the VSB-unaided condition versus 94% for the VSB-aided condition. Conclusions: Although the unilateral VSB middle ear implant does not restore stereophony, it does improve speech intelligibility in background noise.

Evaluation of Speech Intelligibility under Different Lighting Conditions

Evaluation of Speech Intelligibility under Different Lighting Conditions

Issues relevant to the evaluation of speech intelligibility with cochlear implant recipients

In a series of experiments aimed at identifying factors that influence speech perception by cochlear implant (CI) users, we are currently investigating the reception and utilization of relevant perceptual information available in monosyllables and steady‐state vowels. Normal‐hearing subjects and adults fitted with Clarion and Nucleus cochlear implant (CI) devices provided identification responses to both natural and synthetic speech. The first experiment examines intelligibility differences of a standard word identification test (NU‐6) recorded in three different contexts by two talkers (male/female): (1) individual words spoken in typical citation format, and the same words excised from a narrative read at (2) a typical rate and (3) a rapid rate. The second experiment demonstrates the potential diagnostic value afforded from mapping the acoustic vowel space using steady‐state synthetic tokens. The last experiment explores the effects of harmonic richness and first formant (F1) emphasis on synthetic vowel identification by creating tokens varying only in F0, such that a harmonic of F0 would coincide with the F1 center frequency in (1) a male (harmonically rich) F0 and (2) a female (harmonically sparse) F0 range, and (3) an intermediate F0 which was harmonically noncoincident with F1. [Work supported by OSU Office of Research.]

Contact with Students with Hearing Impairments and the Evaluation of Speech Intelligibility and Personal Qualities

This study examined the effect of contact with mainstreamed high school students with hearing impairments on hearing students' evaluations of personal qualities of speakers with three different levels of speech intelligibility (SI) and on their attitudes toward deafness (ATD). Seventy high school students who studied with classmates with hearing impairments (group with contact) and 70 students who did not study with classmates with hearing impairments (group without contact) listened to recordings of three speakers with good, moderate, and poor SI and evaluated the speakers' personal qualities using a semantic differential scale that included emotional-behavioral and cognitive factors. In addition, the participants evaluated the speakers' SI and filled out a questionnaire regarding their ATD. In general, better SI was associated with the attribution of more positive personal qualities. However, students with contact expressed more positive evaluations (than the students without contact) of the speaker with the poor SI on both the emotional-behavioral and the cognitive factors of the personal qualities questionnaire. These evaluations were similar to evaluations of the speaker with moderate SI. Students without contact expressed more positive evaluations of the speaker with the moderate SI on the cognitive factor than did students with contact. Finally, a positive relation was found between ATD and evaluation of personal qualities of the speaker with the poor SI on the cognitive and emotional-behavioral factors.

An Analytical Look at Live and CB Speech Intelligibility in Contemporary Commercial Truck Cabs Based on the Articulation Index

The Federal Highway Administration (FHWA) currently has a hearing requirement for individuals seeking a commercial driver's license for interstate commerce. The existence of such a requirement implies that truck driving is a hearing-critical job, perhaps including speech communication tasks. Indeed, a task analysis performed in an earlier phase of this research confirmed that there are many hearing critical truck driving tasks, including several speech communication tasks. However, previous studies have shown that the truck cab environment can be quite noisy, perhaps noisy enough to impair speech communication. In this study, an analytical method known as the Articulation Index (AI) was used to predict the level of intelligibility in truck cabs for live and CB speech given certain measured truck-cab noise levels. Results showed that except in the quietest noise conditions, speech communication is severely degraded in the truck cab, a finding with potential safety implications. The AI method described herein can be used as a standardized analytical technique to evaluate speech intelligibility in various truck cab environments for design and comparison purposes.

Relationship between speech transmission index and easiness of speech perception in reverberatory fields

When evaluating speech intelligibility in rooms, the Tri-syllable articulation test is used as one of the standard indicators in Japan. Sometimes there are differences between auditory impression and the score of the articulation test. So the psychological scale values of easiness of speech perception (ESP) were measured by experiments, which indicated differences between ESP and articulation scores and a good correlation (R=0.9) between ESP and speech transmission index (STI) in reverberation. The purpose of this paper is to point out relationships between the condition of sound field and ESP. The first of two experiments in this study was done in reverberatory fields. Sound-pressure level (SPL) of direct sound, direct sound to first reflection ratio (D/R), initial delay time, frequency characteristics of reverberation, and reverberation time (RT) were varied. The other experiment was done in simulated fields of actual auditoriums. The result of the first experiment shows D/R and RT has alarge effect on ESP and while varying D/R, the correlation between ESP and STI is not good (R=0.6). The second experiment indicates that when the sound source is on the center of the stage, the correlation between STI and ESP is 0.8; and a different tendency is found while using the sound reinforcement system.

Trial of Objective Evaluation of Speech Intelligibility

Trial of Objective Evaluation of Speech Intelligibility

A pilot study of SPINE test scores and measures of tongue deviancy in speakers with severe-to-profound hearing loss.

Two developments show promise in the assessment and remediation of defective speech production in persons with hearing loss. A perceptual speech-intelligibility test, the SPINE (for Speech Intelligibility Evaluation), is a simple, clinician-administered instrument which is valid, reliable, and clinically efficient. In addition, the development of acoustic measures of tongue deviancy, computed from formant frequencies, makes possible a direct lateral visualization of tongue placement in relation to standard vowel placement. In this study, SPINE test scores of 28 persons with severe-to-profound hearing loss were correlated with two measures of tongue deviancy during production of the vowels /i/, /a/, and /u/. For both measures of tongue deviancy, correlations with the SPINE were significant for the three vowels combined and for the isolated vowel /i/. These findings suggest that clinicians may ultimately have two different but complementary means of assessing speech production in persons with hearing loss.

Comparing Abilities of Children With Profound Hearing Impairments to Learn Consonants Using Electropalatography or Traditional Aural.Oral Techniques

Two groups of nine children with profound hearing impairments and low intelligibility were taught to produce the consonants /t,d,k,g,s,z,S/ using either electropalatographic (palatometry) or traditional aural-oral techniques. Testing was completed pre-, immediately post-, and 6 months post-treatment by examining productions of CV syllables (V = /i,a/) using electropalatography-determined linguapalatal contacts and listener identifications. Intelligibility was also measured using the CID Picture Speech Intelligibility Evaluation (SPINE) test. Both groups improved their consonant productions as a result of 26 50-minute sessions. Sessions were given twice daily over 3- to 4-week training periods. Immediately post-treatment, the electropalatography-trained subjects produced better consonants as measured by linguapalatal contact patterns and listener identifications. The linguapalatal-contact patterns learned by the electropalatography-trained group better matched normal speaker productions than did those of the traditionally trained group. Both groups showed equal improvement for both post-treatment conditions when tested with the CID SPINE test. Although further research is needed, the results of this study suggest that electropalatographic techniques are, at least, equal alternatives to traditional aural-oral speech training techniques for speakers with profound hearing impairments.

Evolution spatiale de l'intelligibilité de la parole dans un habitacle automobile

In a car, speech intelligibility primarily depends on the ratio between speech and noise levels. The speech level, as perceived by the listener, is influenced by the compared positions of the speaker and the listener and the characteristics of the car cabin (its geometry and materials). In this study, the Articulation Index was used to evaluate speech intelligibility in different positions in the compartment of a luxury sedan car, for two sound sources: an omnidirectionnal one and a loudspeaker whose directivity is the same as a human speaker's. The effect of human's directivity is to decrease speech intelligibility, especially in cases where the communication is difficult (from the front seats to the rear ones). These measurements were also compared with simple models of sound field in the car (free field and diffuse field). For an omnidirectional source, the field is nearly a free one; for a directional one, the surfaces of the cabin can reflect sounds to increase the level, but this increase is not equivalent to what can be seen in a diffuse field. Therefore, the geometry and absorption characteristics in a car compartment could be modified in order to improve communication between passengers

Comparison of the speech transmission index and the modified rhyme test in simulated cockpit ambient noise.

The military aircraft test and evaluation community has recently introduced a much quicker, less obtrusive method for evaluating speech intelligibility of a communication system. The method uses the Speech Transmission Index (STI) to predict intelligibility using 15 s of flight time. Previous research has developed a correlation between this method and subjective intelligibility scores for the Phonetically Balanced (PB) word test. This research has not yet shown an experimental relationship to the Modified Rhyme Test (MRT), one preferred method of the military. In addition, the research has not accounted for ambient noise with a spectrum that represents an in-flight aircraft cockpit environment. In the present study, the relationship between the STI and MRT in simulated noise was investigated. Two types of noise were used. The first type was noise with a spectrum to represent in-flight ambient cockpit noise. The second was pink noise (equal energy per octave), a commonly used representation of environmental cockpit noise. The study was two phase. Phase one measured the intelligibility of a communication channel degraded with the noise presented at different signal-to-noise ratios. The intelligibility test was the MRT. Phase two used the same communication channel and signal-to-noise ratio conditions to measure objectively the intelligibility using the STI. The findings of the study showed that the MRT scores were significantly differ ent between the different noise presentations while the MRT scores and STI yielded approximately the same values.