Speech Intelligibility Index Model Research Articles

Effects of Contextual Predictability Clues in Speech Materials on the Korean Speech Intelligibility Index

Purpose: This study aimed to derive band-importance functions (BIFs) and transfer functions (TFs) according to contextual predictability clues to determine the influence of contextual predictability clues in Korean speech material on the speech intelligibility index (SII). Methods: This study was conducted on 156 native speakers of Korean who had normal hearing. Korean speech perception in noise test material, which was composed of 120 high-predictability and 120 low-predictability sentences, was used for stimuli. To obtain intelligibility data, participants were tested for intelligibility in various frequency ranges and signal-to-noise ratio conditions. In order to derive the BIF and the TF, a nonlinear optimization procedure using MATLAB (MathWorks, Inc.) was used. Results: The BIF derived from the high-predictability sentences showed a peak in areas of 700 Hz (7.0%), 1,850 Hz (8.5%), and 4,800 Hz (7.6%). The crossover frequency for the high-predictability sentences was around 1,370 Hz. The BIF derived from the low-predictability sentences showed a peak in areas of 570 Hz (7.5%), 1,850 Hz (9.3%), and 4,000 Hz (8.0%). The crossover frequency for the low-predictability sentences was around 1,600 Hz. In the case of the TF, the TF curves derived from high-predictability sentences were steeper than those derived from low-predictability sentences.Conclusion: In the SII model, speech intelligibility differs according to contextual predictability clues. Especially, the more contextual predictability clues at the identical audibility, the higher the intelligibility predicted by the SII. Therefore, accurate speech intelligibility prediction requires the use of SII considering the contextual predictability clues that are characteristic of the stimulus.

The effect of speech material on the band importance function for Mandarin Chinese.

Speech material influences the relative contributions of different frequency regions to intelligibility for English. In the current study, whether a similar effect of speech material is present for Mandarin Chinese was investigated. Speech recognition was measured using three speech materials in Mandarin, including disyllabic words, nonsense sentences, and meaningful sentences. These materials differed from one another in terms of the amount of contextual information and word frequency. The band importance function (BIF), as defined under the Speech Intelligibility Index (SII) framework, was used to quantify the contributions across frequency regions. The BIFs for the three speech materials were estimated from 16 adults who were native speakers of Mandarin. A Bayesian adaptive procedure was used to efficiently estimate the octave-frequency BIFs for the three materials for each listener. As the amount of contextual information increased, low-frequency bands (e.g., 250 and 500 Hz) became more important for speech recognition, consistent with English. The BIF was flatter for Mandarin than for comparable English speech materials. Introducing the language- and material-specific BIFs to the SII model led to improved predictions of Mandarin speech-recognition performance. Results suggested the necessity of developing material-specific BIFs for Mandarin.

Predicting Speech Recognition Using the Speech Intelligibility Index and Other Variables for Cochlear Implant Users.

Purpose Although the speech intelligibility index (SII) has been widely applied in the field of audiology and other related areas, application of this metric to cochlear implants (CIs) has yet to be investigated. In this study, SIIs for CI users were calculated to investigate whether the SII could be an effective tool for predicting speech perception performance in a population with CI. Method Fifteen pre- and postlingually deafened adults with CI participated. Speech recognition scores were measured using the AzBio sentence lists. CI users also completed questionnaires and performed psychoacoustic (spectral and temporal resolution) and cognitive function (digit span) tests. Obtained SIIs were compared with predicted SIIs using a transfer function curve. Correlation and regression analyses were conducted on perceptual and demographic predictor variables to investigate the association between these factors and speech perception performance. Result Because of the considerably poor hearing and large individual variability in performance, the SII did not predict speech performance for this CI group using the traditional calculation. However, new SII models were developed incorporating predictive factors, which improved the accuracy of SII predictions in listeners with CI. Conclusion Conventional SII models are not appropriate for predicting speech perception scores for CI users. Demographic variables (aided audibility and duration of deafness) and perceptual-cognitive skills (gap detection and auditory digit span outcomes) are needed to improve the use of the SII for listeners with CI. Future studies are needed to improve our CI-corrected SII model by considering additional predictive factors. Supplemental Material https://doi.org/10.23641/asha.8057003.

An Analysis of Individual Differences in Recognizing Monosyllabic Words Under the Speech Intelligibility Index Framework.

Individual differences in the recognition of monosyllabic words, either in isolation (NU6 test) or in sentence context (SPIN test), were investigated under the theoretical framework of the speech intelligibility index (SII). An adaptive psychophysical procedure, namely the quick-band-importance-function procedure, was developed to enable the fitting of the SII model to individual listeners. Using this procedure, the band importance function (i.e., the relative weights of speech information across the spectrum) and the link function relating the SII to recognition scores can be simultaneously estimated while requiring only 200 to 300 trials of testing. Octave-frequency band importance functions and link functions were estimated separately for NU6 and SPIN materials from 30 normal-hearing listeners who were naïve to speech recognition experiments. For each type of speech material, considerable individual differences in the spectral weights were observed in some but not all frequency regions. At frequencies where the greatest intersubject variability was found, the spectral weights were correlated between the two speech materials, suggesting that the variability in spectral weights reflected listener-originated factors.

A Comparison of Gain for Adults from Generic Hearing Aid Prescriptive Methods: Impacts on Predicted Loudness, Frequency Bandwidth, and Speech Intelligibility

Prescriptive methods have been at the core of modern hearing aid fittings for the past several decades. Every decade or so, there have been revisions to existing methods and/or the emergence of new methods that become widely used. In 2001 Byrne et al provided a comparison of insertion gain for generic prescriptive methods available at that time. The purpose of this article was to compare National Acoustic Laboratories-Non-linear 1 (NAL-NL1), National Acoustic Laboratories-Non-linear 2 (NAL-NL2), Desired Sensation Level Multistage Input/Output (DSL m[i/o]), and Cambridge Method for Loudness Equalization 2-High-Frequency (CAMEQ2-HF) prescriptive methods for adults on the amplification characteristics of prescribed insertion gain and compression ratio. Following the differences observed in prescribed insertion gain among the four prescriptive methods, analyses of predicted specific loudness, overall loudness, and bandwidth of cochlear excitation and effective audibility as well as speech intelligibility of the international long-term average speech spectrum (ILTASS) at an average conversational input level were completed. These analyses allow for the discussion of similarities and differences among the present-day prescriptive methods. The impact of insertion gain differences among the methods is examined for seven hypothetical hearing loss configurations using models of loudness perception and speech intelligibility. Hearing loss configurations for adults of various types and degrees were selected, five of which represent sensorineural impairment and were used by Byrne et al; the other two hearing losses provide an example of mixed and conductive impairment. Prescribed insertion gain data were calculated in 1/3-octave frequency bands for each of the seven hearing losses from the software application of each prescriptive method over multiple input levels. The insertion gain data along with a diffuse field-to-eardrum transfer function were used to calculate output levels at the eardrums of the hypothetical listeners. Levels of hearing loss and output were then used in the Moore and Glasberg loudness model and the ANSI S3.5-1997 Speech Intelligibility Index model. NAL-NL2 and DSL m[i/o] provided comparable overall loudness of approximately 8 sones for the five sensorineural hearing losses for a 65 dB SPL ILTASS input. This loudness was notably less than that perceived by a normal-hearing person for the same input signal, 18.6 sones. NAL-NL2 and DSL m[i/o] also provided comparable predicted speech intelligibility in quiet and noise. CAMEQ2-HF provided a greater average loudness, similar to NAL-NL1, with more high-frequency bandwidth but no significant improvement to predicted speech intelligibility. Definite variation in prescribed insertion gain was present among the prescriptive methods. These differences when averaged across the hearing losses were, by and large, negligible with regard to predicted speech intelligibility at normal conversational speech levels. With regard to loudness, DSL m[i/o] and NAL-NL2 provided the least overall loudness, followed by CAMEQ2-HF and NAL-NL1 providing the most loudness. CAMEQ2-HF provided the most audibility at high frequencies; even so, the audibility became less effective for improving speech intelligibility as hearing loss severity increased.

Modelling the speech reception threshold in non-stationary noise in hearing-impaired listeners as a function of level

The extended speech intelligibility index (ESII) model () forms an upgrade to the conventional speech intelligibility index model. For normal-hearing listeners the ESII model is able to predict the speech reception threshold (SRT) in both stationary and non-stationary noise maskers. In this paper, a first attempt is made to evaluate the ESII with SRT data obtained by , and of hearing-impaired listeners in stationary, 10-Hz interrupted, and non-stationary speech-shaped noise measured at different noise levels. The results show that the ESII model is able to describe the SRT in different non-stationary noises for normal-hearing listeners at different noise levels reasonably well. However, the ESII model is less successful in the case of predicting the SRT in non-stationary noise for hearing-impaired subjects. As long as the present audibility models cannot describe the auditory processing in a listener with cochlear hearing loss accurately, it is difficult to distinguish between raised SRTs due to supra-threshold deficits or factors such as cognition, age, and language skills.SumarioEl modelo del Índice extendido de inteligibilidad del lenguaje (ESII) () constituye una actualización del modelo de índice convencional de inteligibilidad del lenguaje. Para oyentes con audición normal, el modelo del ESII es capaz de predecir el umbral de recepción del lenguaje (SRT) tanto ante enmascaradores de ruido estacionario y como no estacionario. En este trabajo, se realiza un primer intento para evaluar el ESII con datos del SRT obtenidos por Laat y Plomp (1983) y a partir de sujetos con hipoacusia, en ruido estacionario, con interrupciones de 10-Hz, y en ruido no estacionario con conformación de lenguaje, medido a diferentes niveles de intensidad. Los resultados muestran que el modelo ESII es capaz de describir el SRT razonablemente bien, para diferentes ruidos no estacionarios en sujetos normo-oyentes, a diferentes niveles de intensidad. Sin embargo, el modelo ESII es menos exitoso para predecir el SRT en ruido no estacionario con sujetos hipoacúsicos. En tanto que los modelos presentes de audibilidad no puedan describir el procesamiento auditivo en un sujeto con pérdida auditiva coclear en forma exacta, es difícil distinguir entre un SRT elevado debido a una deficiencia supra-umbral o a factores tales como cognición, edad o habilidad para el lenguaje.

Modeling speech intelligibility in quiet and noise in listeners with normal and impaired hearing

The speech intelligibility index (SII) is an often used calculation method for estimating the proportion of audible speech in noise. For speech reception thresholds (SRTs), measured in normally hearing listeners using various types of stationary noise, this model predicts a fairly constant speech proportion of about 0.33, necessary for Dutch sentence intelligibility. However, when the SII model is applied for SRTs in quiet, the estimated speech proportions are often higher, and show a larger inter-subject variability, than found for speech in noise near normal speech levels [65 dB sound pressure level (SPL)]. The present model attempts to alleviate this problem by including cochlear compression. It is based on a loudness model for normally hearing and hearing-impaired listeners of Moore and Glasberg [(2004). Hear. Res. 188, 70-88]. It estimates internal excitation levels for speech and noise and then calculates the proportion of speech above noise and threshold using similar spectral weighting as used in the SII. The present model and the standard SII were used to predict SII values in quiet and in stationary noise for normally hearing and hearing-impaired listeners. The present model predicted SIIs for three listener types (normal hearing, noise-induced, and age-induced hearing loss) with markedly less variability than the standard SII.

Frequency importance function of Mandarin Chinese speech

The speech intelligibility index (SII) theory objectively assesses speech intelligibility, and the frequency-importance function (FIF), which reflects the relative importance of various frequency bands to speech intelligibility for various languages, occupies the central part of the theory. However, the FIF has not been examined for tonal Chinese Mandarin speech. In this study, considering the characteristics of Mandarin speech, 50 phonemically-balanced one-syllable words were selected as the speech stimuli from the speech corpus "A Method for Calculating Articulation Index (GB/T 15508-1995)", which meets the National Standards of China, and the 1/3 octave FIF was measured and computed with the Fletcher's method. The results show that spectrum regions with frequencies 2000-4000 Hz are more important in Chinese speech than those in English speech. The results predicted by SII model with the new FIF fit the result of human psychophysical studies well, which indicate that the new FIF model is effective and appropriate for assessing the intelligibility of Mandarin speech even when a steady-state noise masker is co-presented. [Supported by the NSFC 60435010; 60535030; 60605016; 30670704]

A model with compression for the estimation of speech intelligibility in quiet

The Speech Intelligibility Index (SII) is an often used calculation method for estimating the proportion of speech information available in noise. For speech reception thresholds (SRTs), measured in normal-hearing listeners using various types of stationary noise, this model predicts a fairly constant speech proportion (about 0.30) necessary for sentence intelligibility. However, when the SII model is applied for SRTs in quiet, the estimated speech proportions are often lower, and show a larger inter-subject variability, than found for speech in noise near the normal speech level (65 dB SPL). The present model attempts to alleviate this problem by including cochlear compression. It is based on a loudness model for normal-hearing and hearing-impaired listeners of Moore and Glasberg [Hear. Res. 188, 70–88 (2004)]. It estimates internal excitation levels of the speech and then calculates the proportion of speech above threshold using similar spectral weighting as used in the standard SII. The present model and the standard SII were used to predict SRTs in quiet for both normal-hearing and hearing-impaired listeners. The present model predicted a set of data for three listener types (normal hearing, noise-induced hearing loss, and age-induced hearing loss) with less variability than the standard SII.

Noise-management algorithm may improve speech intelligibility in noise

A likely consequence of sensorineural hearing loss is a diminished ability to understand speech in noisy backgrounds. Indeed, the inability to hear in noise is one of the main reasons for dissatisfaction with hearing aid use.1 Although hearing aids with a directional microphone provide substantial improvement in the ability of wearers to understand speech in noisy environments,2 space requirements and the uncertainty of directional characteristics when a directional hearing aid is inserted deeply into the ear canal prevent the implementation of such technology in the smallest completely-in-the-canal (CIC) hearing aids. A patient who insists on wearing CIC hearing aids will have to rely solely on the efficacy of single-mic noise-management strategies. Currently, most commercial “noise reduction” schemes use the modulation rates of the input signal as a basis for estimating the “speech” and “noise” nature of the input. “Speech” sounds are amplified with input-dependent gain, while “noise” sounds are typically amplified at a reduced gain level beyond input-dependent levels (see Chung, 20043). The exact amount of gain reduction and its time course vary greatly among manufacturers. Despite such differences, most studies have reported only an improvement in listening comfort by use of such a scheme.4 The improvement in listening comfort is noteworthy. It suggests that hearing-impaired persons whose hearing aids have noise reduction may be less affected by high-output sound pressure levels, less stressed, and less distracted in noisy situations, and may be more likely to attend to the sound sources and wear their hearing aids longer than if they lacked this feature. Wearers may also improve their participation in daily activities and their quality of life. It is possible that the reduction in stress level (or distractions) may improve the user’s speech understanding in some noisy environments.

An SII-based approach to predict the speech intelligibility in fluctuating noise for normal-hearing listeners

The speech intelligibility index (SII) is frequently used to predict the speech intelligibility for speech in a given interfering noise. However, the SII model only has been validated for speech in stationary noise. Since the SII departs from speech and noise spectra, it does not take into account any fluctuations in the masking noise. Hence, the model will yield similar SII values, regardless of the degree of fluctuation. In contrast, from the literature it is clear that normal-hearing listeners can benefit from the fluctuations in the noise. The present paper describes an SII-based approach to model speech reception thresholds (SRTs) for speech in both stationary and fluctuating noise. The basic principle of this approach is that both speech and noise signals are partitioned into small time frames. Within each time frame, the conventional SII is determined, yielding the speech information available to the listener at that time frame. Next, the SII values of these time frames are averaged, resulting in the SII for that particular condition. With the aid of SRT data from the literature, it will be shown that this approach can give a good account for most existing data.

Maximizing effective audibility in hearing aid fitting.

This paper examines why more audibility is not always better than less audibility if hearing-impaired people are to best understand speech. We used speech perception data from 14 normally hearing and 40 hearing-impaired people to quantify the contribution of audibility to speech intelligibility. The quantification revealed that the effectiveness of audibility decreased with hearing loss, and the decrement was greater at high frequencies than at lower frequencies. To apply the Speech Intelligibility Index (SII) model to predict speech intelligibility for hearing-impaired people, we modified the model to take account of effective audibility rather than physical audibility. The modified SII model provided an adequate description of speech performance of people with a wide range of hearing threshold levels. We applied the model to the evaluation of two prescriptions for a sloping audiogram at prescribed levels and at equated loudness levels to demonstrate the necessity of considering loudness and effective audibility in prescribing amplification. Effective audibility is defined as audibility corrected for the effects of level distortion and hearing loss desensitization, and this paper proposes a method of estimating effective audibility from hearing threshold level at different frequencies. The practical implication of considering effective audibility in prescribing hearing aids is that for a given listening level, less gain is provided at frequencies where the hearing is most impaired to allow more gain at frequencies where audibility is most useful. In developing the NAL-NL1 prescription for nonlinear hearing aids, we adopted the modified SII model together with a loudness model to derive optimal gain-frequency response characteristics that maximize predicted speech intelligibility for people with different degrees of hearing losses.

The effect of intensity perturbations on speech intelligibility for normal-hearing and hearing-impaired listeners.

Hearing-impaired listeners are known to suffer from reduced speech intelligibility in noise, even if sounds are above their hearing thresholds. This study examined the possible contribution of reduced acuity of intensity coding to this problem. The "distortion-sensitivity model" was used: the effect of reduced acuity of auditory intensity coding on intelligibility was mimicked by an artificial distortion of the speech intensity coding, and the sensitivity to this distortion for hearing-impaired listeners was compared with that for normal-hearing listeners. Stimuli (speech plus noise) were wavelet coded using a Gaussian wavelet (1/4-octave bandwidth). The intensity coding was distorted by multiplying the modulus of each wavelet coefficient by a random factor. Speech-reception thresholds (SRTs) were measured for various degrees of intensity perturbation. Hearing-impaired listeners were classified as suffering from suprathreshold deficits if intelligibility of undistorted speech was worse than predicted from audibility by the speech intelligibility index model [ANSI, ANSI S3.5-1997 (1997)]. Hearing-impaired listeners without suprathreshold deficits were as sensitive to the intensity distortion as the normal-hearing listeners. Hearing-impaired listeners with suprathreshold deficits appeared to be less sensitive. Results indicate that reduced acuity of auditory intensity coding may be a factor underlying reduced speech intelligibility in noise for the hearing impaired.

Relations between intelligibility of narrow-band speech and auditory functions, both in the 1-kHz frequency region.

Relations between perception of suprathreshold speech and auditory functions were examined in 24 hearing-impaired listeners and 12 normal-hearing listeners. The speech intelligibility index (SII) was used to account for audibility. The auditory functions included detection efficiency, temporal and spectral resolution, temporal and spectral integration, and discrimination of intensity, frequency, rhythm, and spectro-temporal shape. All auditory functions were measured at 1 kHz. Speech intelligibility was assessed with the speech-reception threshold (SRT) in quiet and in noise, and with the speech-reception bandwidth threshold (SRBT), previously developed for investigating speech perception in a limited frequency region around 1 kHz. The results showed that the elevated SRT in quiet could be explained on the basis of audibility. Audibility could only partly account for the elevated SRT values in noise and the deviant SRBT values, suggesting that suprathreshold deficits affected intelligibility in these conditions. SII predictions for the SRBT improved significantly by including the individually measured upward spread of masking in the SII model. Reduced spectral resolution, reduced temporal resolution, and reduced frequency discrimination appeared to be related to speech perception deficits. Loss of peripheral compression appeared to have the smallest effect on the intelligibility of suprathreshold speech.

Measuring the threshold for speech reception by adaptive variation of the signal bandwidth. II. Hearing-impaired listeners.

In a previous study [Noordhoek et al., J. Acoust. Soc. Am. 105, 2895-2902 (1999)], an adaptive test was developed to determine the speech-reception bandwidth threshold (SRBT), i.e., the width of a speech band around 1 kHz required for a 50% intelligibility score. In this test, the band-filtered speech is presented in complementary bandstop-filtered noise. In the present study, the performance of 34 hearing-impaired listeners was measured on this SRBT test and on more common SRT (speech-reception threshold) tests, namely the SRT in quiet, the standard SRT in noise (standard speech spectrum), and the spectrally adapted SRT in noise (fitted to the individual's dynamic range). The aim was to investigate to what extent the performance on these tests could be explained simply from audibility, as estimated with the SII (speech intelligibility index) model, or require the assumption of suprathreshold deficits. For most listeners, an elevated SRT in quiet or an elevated standard SRT in noise could be explained on the basis of audibility. For the spectrally adapted SRT in noise, and especially for the SRBT, the data of most listeners could not be explained from audibility, suggesting that the effects of suprathreshold deficits may be present. Possibly, such a deficit is an increased downward spread of masking.