The Extended Speech Transmission Index: Predicting speech intelligibility in fluctuating noise and reverberant rooms

A context-based model to predict the intelligibility of sentences in non-stationary noises

The context-based Extended Speech Transmission Index (cESTI) (van Schoonhoven et al., 2022, J. Acoust. Soc. Am. 151, 1404-1415) was successfully applied to predict the intelligibility of monosyllabic words with different degrees of context in interrupted noise. The current study aimed to use the same model for the prediction of sentence intelligibility in different types of non-stationary noise. The necessary context factors and transfer functions were based on values found in existing literature. The cESTI performed similar to or better than the original ESTI when noise had speech-like characteristics. We hypothesize that the remaining inaccuracies in model predictions can be attributed to the limits of the modelling approach with regard to mechanisms, such as modulation masking and informational masking.

A context-based approach to predict intelligibility of meaningful and nonsense words in interrupted noise: Model evaluation

The context-based Extended Speech Transmission Index (cESTI) by Van Schoonhoven et al. (2022) was successfully used to predict the intelligibility of meaningful, monosyllabic words in interrupted noise. However, it is not clear how the model behaves when using different degrees of context. In the current paper, intelligibility of meaningful and nonsense CVC words in stationary and interrupted noise was measured in fourteen normally hearing adults. Intelligibility of nonsense words in interrupted noise at -18 dB SNR was relatively poor, possibly because listeners did not profit from coarticulatory cues as they did in stationary noise. With 75% of the total variance explained, the cESTI model performed better than the original ESTI model (R2 = 27%), especially due to better predictions at low interruption rates. However, predictions for meaningful word scores were relatively poor (R2 = 38%), mainly due to remaining inaccuracies at interruption rates below 4 Hz and a large effect of forward masking. Adjusting parameters of the forward masking function improved the accuracy of the model to a total explained variance of 83%, while the predicted power of previously published cESTI data remained similar.

Spectral weighting for sentence recognition in steady-state and amplitude-modulated noise

Spectral weights in octave-frequency bands from 0.25 to 4 kHz were estimated for speech-in-noise recognition using two sentence materials (i.e., the IEEE and AzBio sentences). The masking noise was either unmodulated or sinusoidally amplitude-modulated at 8 Hz. The estimated spectral weights did not vary significantly across two test sessions and were similar for the two sentence materials. Amplitude-modulating the masker increased the weight at 2 kHz and decreased the weight at 0.25 kHz, which may support an upward shift in spectral weights for temporally fluctuating maskers.

A context-based approach to predict speech intelligibility in interrupted noise: Model design

The Extended Speech Transmission Index (ESTI) by van Schoonhoven et al. [(2019). J. Acoust. Soc. Am. 145, 1178-1194] was used successfully to predict intelligibility of sentences in fluctuating background noise. However, prediction accuracy was poor when the modulation frequency of the masker was low (<8 Hz). In the current paper, the ESTI was calculated per phoneme to estimate phoneme intelligibility. In the next step, the ESTI model was combined with one of two context models {Boothroyd and Nittrouer, [(1988). J. Acoust. Soc. Am. 84, 101-114]; Bronkhorst et al., [(1993). J. Acoust. Soc. Am. 93, 499-509} in order to improve model predictions. This approach was validated using interrupted speech data, after which it was used to predict speech intelligibility of words in interrupted noise. Model predictions improved using this new method, especially for maskers with interruption rates below 5 Hz. Calculating the ESTI at phoneme level combined with a context model is therefore a viable option to improve prediction accuracy.

The relation between the intelligibility of irrelevant speech and cognitive performance-A revised model based on laboratory studies

Irrelevant background speech causes dissatisfaction and impairs cognitive performance in open-plan offices. The model of Hongisto (2005, Indoor Air, 15, 458-468) predicts the relation between cognitive performance and the intelligibility of speech described with an objectively measured quantity, the Speech Transmission Index (STI). The model has impacted research in psychology and room acoustics as well as the acoustic design guidelines of offices. However, the model was based on scarce empirical data. The aim of this study was to revise the model based on a systematic literature review, focusing on laboratory experiments manipulating the STI of speech by wide-band steady-state noise. Fourteen studies reporting altogether 34 tests of the STI-performance relation were included. According to Model 1 that includes all tests, performance begins to decrease approximately above STI = 0.21 while the maximum decrease is reached at STI = 0.44. Verbal short-term memory tasks were most strongly and very consistently affected by the STI of speech. The model for these tasks showed a deterioration in performance between STI 0.12 and 0.51. Some evidence of an STI-performance relation was found in verbal working memory tasks and limited evidence in complex verbal tasks. Further research is warranted, particularly concerning task-specific effects.

Impact of Background Noise Fluctuation and Reverberation on Response Time in a Speech Reception Task

Purpose This study examines the effects of reverberation and noise fluctuation on the response time (RT) to the auditory stimuli in a speech reception task. Method The speech reception task was presented to 76 young adults with normal hearing in 3 simulated listening conditions (1 anechoic, 2 reverberant). Speechlike stationary and fluctuating noise were used as maskers, in a wide range of signal-to-noise ratios. The speech-in-noise tests were presented in a closed-set format; data on speech intelligibility and RT (time elapsed from the offset of the auditory stimulus to the response selection) were collected. A slowing down in RTs was interpreted as an increase in listening effort. Results RTs slowed down in the more challenging signal-to-noise ratios, with increasing reverberation and for stationary compared to fluctuating noise, consistently with a fluctuating masking release scheme. When speech intelligibility was fixed, it was found that the estimated RTs were similar or faster for stationary compared to fluctuating noise, depending on the amount of reverberation. Conclusions The current findings add to the literature on listening effort for listeners with normal hearing by indicating that the addition of reverberation to fluctuating noise increases RT in a speech reception task. The results support the importance of integrating noise and reverberation to provide accurate predictors of real-world performance in clinical settings.