Intelligibility Of Degraded Speech Research Articles

The speech signal enhancement approach with multiple sub-frames analysis for complex magnitude and phase spectrum recompense

The intended speech must be dealt with in the process of speech communication while under the impact of noise experienced in a variety of situations that degrade speech intelligibility and quality. This work proposes a multiple sub-frames analysis for the elimination of noise variants with compensation of the magnitude and phase spectrum from speech degraded by noise. The clean speech samples are extracted from the ITU-T recommended dataset at a 16 kHz sampling rate and down-sampled to an 8 kHz sampling rate. The noise signal variants are added from the AURORA and NOISEX-92 datasets at diverse input SNR levels (0 dB, 5 dB, 10 dB, 15 dB). The duration of window frames is chosen to be 25 msec in length, together with a shift percentage of 40%, to maintain the continuous nature of frames in speech. The smoothing factor for noise updating in a specific sub-frame is set to 9, and the spectral floor parameter for determining the precise amount of noise elimination in the corrupted spectrum is set to 0.03. The phase spectrum is compensated by incorporating a recompense function that is updated in combination with the sub-frame analysis. The accomplishment of the suggested approach is assessed with regard to objective metrics, including Segmental-Signal-to-Noise-Ratio (SegSNR), Mean-Square-Error (MSE), and Perceptual-Evaluation-of-Speech-Quality (PESQ) scores corresponding to specific sub-frames of speech, respectively. To further analyze the improved quality, simple listening assessment and spectrogram analysis are incorporated, followed by a comparative investigation with prior noise-suppressive algorithms on the corrupted speech corpus.

Effects of urgent speech and congruent/incongruent text on speech intelligibility for older adults in the presence of noise and reverberation

Public address (PA) announcements are widely used, but noise and reverberation can render them unintelligible, and such an environment tends to degrade speech intelligibility more for older adults (OAs) than for younger adults (YAs). Furthermore, in an emergency, textual information available to the public may not coincide with PA announcements, and this mismatch may also contribute to degrading the intelligibility of the announcement. In this study, speech spoken in a normal or urgent style and preceded by congruent or incongruent text was presented to OA participants to investigate the effects of these parameters on word intelligibility in the presence of babble noise and reverberation. This study also investigated age-related differences between OAs and YAs using the same stimuli. The results showed that, while YAs performed better than OAs, the effect of urgent speech did not significantly differ with age. Urgent speech was more intelligible than normal speech, and the congruent text condition was more intelligible than the incongruent and no text conditions. An exploratory analysis showed that urgent speech improved speech intelligibility for OAs when congruent text was available, but not when absent or incongruent. Without congruent information, this listening test may have been too demanding for OAs, who could have had more difficulty processing the fixed timeline of the listening test, the faster urgent speech relative to normal speech, and informational masking because of the slowing of cognitive processing with age. A correlation was found between the average word correct rate for urgent speech and average audiometric thresholds; on the other hand, increased fundamental frequencies in urgent compared with normal speech was not the main predictor of the intelligibility of urgent speech among the OAs. The congruent text benefit in urgent speech was larger for OAs than for YAs in the previous study using the same stimuli. This demonstrates that compared with YAs, OAs may rely more on supportive prior knowledge or receive more benefit from the preceding text in challenging listening environments. These findings imply that age-related cognitive decline may explain the difficulties in speech perception in the presence of noise and reverberation among OAs; however, no definitive conclusions can be drawn. These results suggest that simple combinations of speaking style and textual information affect the intelligibility of emergency PA announcements among OAs, and thus, audiovisual congruence should be considered when announcements are made in public spaces.

Effect of reverberation and noise type on speech intelligibility in real complex acoustic scenarios

Enhanced speech intelligibility in realistic acoustic scenarios guarantees effective communication. Noise and reverberation degrade speech intelligibility, but the combined effect of reverberation, informational noise and position of target, listener and noise source on speech intelligibility still needs to insights. This work investigates the effect of real complex acoustic scenarios on speech intelligibility for adults. In two primary-school classrooms with reverberation time of 0.4 s and 3.1 s, receivers were located on axis with the target at increasing distances, and noise sources were both co-located and spatially separated from the target at different distances providing various degrees of energetic and informational masking. Noise level was maintained constant and set at 60 dB at the receiver's position, regardless of the noise source distance. The longer reverberation time resulted in worse speech recognition thresholds (SRT80s) by 6 dB on average. The competitive effect of informational noise resulted in increased SRT80s by 7 dB on average, in comparison to energetic noise. The SRT80 increases by ~2 dB with doubling the target-to-receiver distance when the noise source is close to the receiver, accounting for both reverberations and noise types. The spatial release from masking resulted in improved speech intelligibility by up to ~3 dB when the noise source is 1 m far from the receiver for energetic masker for low reverberation and, unexpectedly, for the informational masker in high reverberation. This may indicate that a perceptual segregation mechanism sorts out competing voices of informational masker according to their directions in least favorable listening situations.

The Effect of Hearing-Protection Devices on Auditory Situational Awareness and Listening Effort.

Hearing-protection devices (HPDs) are made available, and often are required, for industrial use as well as military training exercises and operational duties. However, these devices often are disliked, and consequently not worn, in part because they compromise situational awareness through reduced sound detection and localization performance as well as degraded speech intelligibility. In this study, we carried out a series of tests, involving normal-hearing subjects and multiple background-noise conditions, designed to evaluate the performance of four HPDs in terms of their modifications of auditory-detection thresholds, sound-localization accuracy, and speech intelligibility. In addition, we assessed their impact on listening effort to understand how the additional effort required to perceive and process auditory signals while wearing an HPD reduces available cognitive resources for other tasks. Thirteen normal-hearing subjects participated in a protocol, which included auditory tasks designed to measure detection and localization performance, speech intelligibility, and cognitive load. Each participant repeated the battery of tests with unoccluded ears and four hearing protectors, two active (electronic) and two passive. The tasks were performed both in quiet and in background noise. Our findings indicate that, in variable degrees, all of the tested HPDs induce performance degradation on most of the conducted tasks as compared to the open ear. Of particular note in this study is the finding of increased cognitive load or listening effort, as measured by visual reaction time, for some hearing protectors during a dual-task, which added working-memory demands to the speech-intelligibility task. These results indicate that situational awareness can vary greatly across the spectrum of HPDs, and that listening effort is another aspect of performance that should be considered in future studies. The increased listening effort induced by hearing protectors may lead to earlier cognitive fatigue in noisy environments. Further study is required to characterize how auditory performance is limited by the combination of hearing impairment and the use of HPDs, and how the effects of such limitations can be linked to safe and effective use of hearing protection to maximize job performance.

Influence of Classroom Acoustics on Noise Disturbance and Well-Being for First Graders.

Several studies have shown so far that poor acoustics inside classrooms negatively affects the teaching and learning processes, especially at the lowest grades of education. However, the extent to which noise exposure or excessive reverberation affect well-being of children at school in their early childhood is still unanswered, as well as their awareness of noise disturbance. This work is a pilot study to investigate to which extent classroom acoustics affects the perceived well-being and noise disturbance in first graders. About 330 pupils aged from 6 to 7 years participated in the study. They belonged to 20 classes of 10 primary schools located in Torino (Italy), where room acoustic measurements were performed and where noise level was monitored during classes. The school buildings and the classrooms were balanced between socioeconomic status and acoustic conditions. Trained experimenters administered questionnaires in each class, where pupils answered all together during the last month of the school year (May). Questions included the happiness scale, subscales assessing self-esteem, emotional health, relationship at home and with friends, enjoyment of school, intensity and noise disturbance due to different sound sources, and quality of voice. The findings of the study suggest that long reverberation times, which are associated with poor classroom acoustics as they generate higher noise levels and degraded speech intelligibility, bring pupils to a reduced perception of having fun and being happy with themselves. Furthermore, bad classroom acoustics is also related to an increased perception of noise intensity and disturbance, particularly in the case of traffic noise and noise from adjacent school environments. Finally, happy pupils reported a higher perception of noise disturbance under bad classroom acoustic conditions, whereas unhappy pupils only reported complaints in bad classroom acoustics with respect to the perception of pleasances with himself or herself and of fitting in at school. Being a mother tongue speaker is a characteristic of children that brings more chances of attending classes in good acoustics, of being less disturbed, and of having more well-being, and richer districts presented better acoustic conditions, in turn resulting in richer districts also revealing a greater perception of well-being.

A speech-based computational auditory signal processing and perception model.

A new speech intelligibility prediction model is presented which is based on the Computational Auditory Signal Processing and Perception model (CASP) of Jepsen, Ewert, and Dau [(2008). J. Acoust. Soc. Am. 124(1), 422-438]. The model combines a non-linear auditory-inspired preprocessing with a backend based on the cross-correlation between the clean and the degraded speech representations in the modulation envelope domain. Several speech degradation and speech enhancement algorithms were considered to study the ability of the model to predict data from normal-hearing listeners. Degradations of speech intelligibility due to additive noise, phase-jitter distortion, and single-channel noise reduction as well as improved speech intelligibility due to ideal binary mask processing are shown to be successfully accounted for by the model. Furthermore, the model reflects stimulus-level dependent effects of auditory perception, including audibility limitations at low levels and degraded speech intelligibility at high levels. Given its realistic non-linear auditory processing frontend, the speech-based computational auditory signal processing and perception model may provide a valuable computational framework for studying the effects of sensorineural hearing impairment on speech intelligibility.

Improving speech intelligibility performance for aircraft maintainers by selecting an appropriate combination of hearing protection devices

Aircraft maintainers work in hazardous noise environments throughout their career. Hazardous noise levels on the flight line vary, affecting the level of hearing protection required to reduce their risk of hearing loss and other hearing related disabilities. Current practice is to provide double hearing protection: a circumaural headset/earmuff with foam earplugs. This combination may provide an adequate amount of protection from aircraft noise, but the proper use of foam earplugs may degrade speech intelligibility (SI) performance. Due to this degradation, many maintainers fit the foam earplugs incorrectly in order to improve SI performance, therefore lowering the amount of protection provided. The objective of this study was to assess the noise attenuation and SI performance of a headset and filtered earplugs worn alone and in combination to provide a more appropriate recommendation of hearing protection devices. The filtered earplug can provide suitable levels of protection when worn alone or in combination with a headset (depending on the noise environment) and can provide adequate “hear through” capability for acceptable SI performance when communicating face-to-face (single hearing protection) or through the aircraft communication system (double hearing protection).

A deep learning based segregation algorithm to increase speech intelligibility for hearing-impaired listeners in reverberant-noisy conditions.

Recently, deep learning based speech segregation has been shown to improve human speech intelligibility in noisy environments. However, one important factor not yet considered is room reverberation, which characterizes typical daily environments. The combination of reverberation and background noise can severely degrade speech intelligibility for hearing-impaired (HI) listeners. In the current study, a deep learning based time-frequency masking algorithm was proposed to address both room reverberation and background noise. Specifically, a deep neural network was trained to estimate the ideal ratio mask, where anechoic-clean speech was considered as the desired signal. Intelligibility testing was conducted under reverberant-noisy conditions with reverberation time T 60 = 0.6 s, plus speech-shaped noise or babble noise at various signal-to-noise ratios. The experiments demonstrated that substantial speech intelligibility improvements were obtained for HI listeners. The algorithm was also somewhat beneficial for normal-hearing (NH) listeners. In addition, sentence intelligibility scores for HI listeners with algorithm processing approached or matched those of young-adult NH listeners without processing. The current study represents a step toward deploying deep learning algorithms to help the speech understanding of HI listeners in everyday conditions.

Speech intelligibility studies in a historic multipurpose room

The Speech-Language-Hearing Sciences department at the University of Minnesota is located in Shevlin Hall, a historic building dating from the 1920s. Shevlin Hall has a large room on the first floor with a high coffered ceiling that is often used for receptions and presentations but also serves as a classroom. Students with hearing loss often complained about attending classes in Shevlin 110 due to its high reverberation times, combined with window AC unit noise, resulting in very degraded speech intelligibility. In 2013, the room went through renovations that addressed room acoustics, sound system design, as well as lighting and equipment (not including AC). Post-renovation measurements and subjective impressions have shown a significant improvement in speech intelligibility in the room. For this paper, we are looking at subject-based intelligibility tests to compare with prediction metrics and simulations, as well as collecting a qualitative narrative on the department’s impression on the room before and after renovations.

Target-word identification in noise with formant enhancement for hearing-impaired listeners

Degraded speech intelligibility in background noise is a common complaint of most listeners with hearing loss. The purpose of the current study is to investigate the effect of spectral enhancement of F2 on target-word identification in noise for the old listeners with hearing loss (HL) and with normal hearing (NH). Target words (e.g., name, color, and digit) were selected and presented based on the paradigm of coordinate response measure (CRM) corpus of English and Chinese versions. Identification tasks with original and F2-enhanced speech in the 2-talker and 6-talker babbles were designed for the two groups (HL and NH). Thresholds of word identification was measured by using adaptive up-down methods. Results showed that listeners with NH had better performance on word identification in noise than listeners with HL in almost all listening conditions. More importantly, thresholds of both groups were improved for enhanced speech signals. Compared with NH group, listeners with HL gained significantly greater benefits in the most challenging condition. These preliminary results suggested that speech sounds with F2 enhancement might improve listeners’ ordinary speech perception in noise.

Downstream changes in firing regularity following damage to the early auditory system

We demonstrate how an abstract mathematical model that approximates a wide range of more detailed models can be used to make predictions about hearing loss-related changes in neural behaviour. One consequence of neurosensory hearing loss (noise-induced and aging-related) is a reduced ability to understand speech, particularly in noisy environments, and sometimes beyond what would be predicted from reduced audibility. Indeed, this type of speech deficit can occur in listeners with near-normal hearing thresholds [1]. A promising avenue of investigation to explain this comes from experimental results in mice showing that there can be a permanent loss of auditory nerve fibres (ANFs) following temporary noise-induced hearing loss (i.e. when thresholds return to normal after a few weeks) [2]. The downstream consequences of this loss of fibres has not yet been systematically investigated (although see [3]). We predict, using a theoretical analysis that applies to a wide range of neural models, that the regularity of the spike trains of many neurons in the cochlear nucleus (the next structure after the auditory nerve) will decrease following a reduction in the number of input cells. We present a mathematical analysis of the stationary behaviour of cells in the ventral cochlear nucleus, approximating them by a stochastic process that is entirely characterised by its mean, standard deviation and time constants. Furthermore, these constants can be straightforwardly related to physiologically significant parameters including the number of inputs and their average firing rates. From this approximation, we can compute the regularity of the chopper cell spike trains measured as the coefficient of variation of their interspike intervals (CV). One simple prediction of this model is that when the intensity of a stimulus changes, leading to a change in the average firing rate of the ANF inputs, there will be a corresponding change in the regularity of the chopper cell spike train. This prediction poses problems for the widely used scheme for classifying chopper cells as sustained or transient based on their ongoing CVs as it implies that the classification could be level-dependent. We present a re-analysis of an existing experimental data set that demonstrates that ongoing CV is indeed level-dependent in the majority of chopper cells, and that in some cells (>7%) this leads to a level-dependence in their classification. Assuming a homeostatic regulation of long term firing rates, a loss of ANFs will lead to an increase in the standard deviation of the stochastic process and a consequent increase in the CV of the chopper cell. Some choppers that were previously classified as sustained will become transient, a substantial change in their behaviour that is highly likely to disrupt auditory processing. While the function of chopper cells is still debated, one suggested role is in the coding of temporal envelope [4], which is widely agreed to be essential for understanding speech. Loss of ANFs could therefore lead to a disruption of the processing of temporal envelope, and consequently degrade speech intelligibility. We briefly conclude by discussing the challenges of testing this hypothesis experimentally.

Speech Intelligibility While Wearing Full-Facepiece Air-Purifying Respirators

Intelligible speech communication while wearing air-purifying respirators is critical for law enforcement officers, particularly when they are communicating with each other or the public. The National Institute for Occupational Safety and Health (NIOSH) requires a 70% overall performance rating to pass speech intelligibility certification for commercial chemical, biological, radiological, and nuclear air-purifying respirators. However, the speech intelligibility of certified respirators is not reported and the impact on operational performance is unknown. The objective of this effort was to assess the speech intelligibility of 12 certified air-purifying respirators and to predict their impact on operational performance. The NIOSH respirator certification standard testing procedures were followed. Regression equations were fit to data from studies that examined the impact of degraded speech intelligibility on operational performance of simple and complex missions. The impact of the tested respirators on operational performance was estimated from these equations. Performance ratings observed for each respirator were: MSA Millennium (90%), 3M FR-M40 (88%), MSA Ultra Elite (87%), Scott M110 (86%), North 5400 (85%), Scott M120 (85%), Avon C50 (84%), Avon FM12 (84%), Survivair Optifit (81%), Drager CDR 4500 (81%), Peltor-AOSafety M-TAC (79%), and 3M FR-7800B (78%). The Millennium and FR-M40 had statistically significantly higher scores than the FR-7800B. The Millennium also scored significantly higher than the M-TAC. All of the tested respirators were predicted to have little impact on simple and complex mission performance times and on simple mission success rate. However, the regression equations showed that 75% of missions that require complex communications would be completed while wearing the Millennium, FR-M40, or Ultra Elite but that only 60% would be completed successfully while wearing the FR-7800B. These results suggest that some certified respirators may have a greater impact on speech communications than others.

GSM interference cancellation for forensic audio: a report on work in progress

A central aspect of forensic phonetic casework concerns the transcription of noisy recordings. An increasing problem in this area of work is the contamination of recordings with interference caused by radio transmissions from GSM mobile phones. Transmitting phones emit short duration radio-frequency pulses at a rate of 217 Hz. The induced interference signal contains the 217 Hz fundamental and a large number of harmonics that overlap the frequency range of speech, and therefore severely degrade speech intelligibility. Listener fatigue is increased due to the harsh sound of theinterference, and overall the transcription of such audio samples is problematic. This paper describes the progressing development of a filter to assist the forensic phonetician in carrying out the transcription of such contaminated recordings

Perceptual consequences of spectro-temporal smearing in the wavelet coding of speech

In this study, speech is spectro-temporally smeared, i.e., the contrasts in the spectral and/or temporal domain are reduced. This spread-of-excitation type of manipulation is embedded in a spectro-temporal analysis and resynthesis scheme. In order to perform a perceptually relevant analysis, wavelet coding with a Gaussian wavelet of 1/4 octave is used. The detectability of smearing and the effect on intelligibility are measured for normal-hearing and hearing-impaired listeners. Smearing is imposed concentrating on the spectral domain (related to upward and downward spread of masking), concentrating on the temporal domain (related to forward and backward masking), and considering the spectral and temporal domain in combination (spectro-temporal spread of excitation). Mechanisms underlying the just-noticeable degree of smearing, and the intelligibility of smeared speech will be addressed. The results may provide insight into the sharpness of the internal spectro-temporal representation of the auditory system and its effect on speech intelligibility. Comparing normal-hearing and hearing-impaired listeners, the role of a reduced sharpness in auditory spectro-temporal coding as a possible explanation for degraded speech intelligibility will be discussed. [Work supported by the Netherlands Organization for Scientific Research (NWO).]

Effects of Speech Intelligibility, Morphological Confusions, and Redundancy on Task Performance

Speech is a vital means of communication for completing many tasks. The speech intelligibility needed for successful communication may be degraded by ambient noise levels, poor communication equipment, or hearing impairments. The present research tested the impact of speech message content on task performance under conditions of degraded speech intelligibility. Sixteen subjects participated in a laboratory experiment using on a board game. Message redundancy, morphological confusions, and speech intelligibility were varied. Task performance and subjective workload were measured. Morphological confusions adversely affected performance (both speed and accuracy) and subjective workload (SWAT) to an increasingly greater extent as speech intelligibility decreased. High redundancy improved the accuracy of performance and subjective workload when speech intelligibility decreased; however, high redundant messages were longer and required more time to process than low redundant messages. These results extend earlier work which measured the impact of morphological confusions and redundancy on speech intelligibility itself to the measurement of their impact on performance. Implications for the development of message content guidelines and their impact on performance are discussed.

Modeling Operational Criteria for Evaluating Speech Communication

Thesis: Evaluation of systems can be conducted best under controlled circumstances which approximate operational conditions. In the present paper, we have examined this thesis as it applies to the test and evaluation of multi-person systems. Our specific focus has been the development of a conceptual model of speech communication requirements and the study of the impact of degraded speech intelligibility on performance of these multi-person systems. To obtain the control necessary to evaluate performance using speech intelligibility, an electronic circuit was developed and employed in a series of simulated tests of operational tasks. The application of these research findings to the task of system test and evaluation is described in the present paper.

An investigation of the effects of hearing protectors on vocal communication in noise : Journal of Sound and Vibration 1975, 41.2 181–196

Previous research has indicated that the wearing of hearing protectors in noise above about 85 dB does not reduce speech intelligibility compared to the open ear condition. However, in spite of this it is still reported by many industrial workers that verbal communication is impeded by such devices being worn in the working environment. Two experiments were carried out to investigate the effects of both earplugs and earmuffs on speech intelligibility for the listener and on the voice level of the talker in the presence of broadband masking noise. It was found that hearing protectors do not degrade speech intelligibility for the listener and may even effect a slight improvement. However, if worn by the talker, his speech is reduced in level and quality to such an extent that intelligibility for the listener is also reduced. When hearing protectors are worn by both talker and listener, the composite effect is an overall reduction in speech intelligibility compared to the open ear condition.

An investigation of the effects of hearing protectors on vocal communication in noise

Previous research has indicated that the wearing of hearing protectors in noise above about 85 dB does not reduce speech intelligibility compared to the open ear condition. However, in spite of this it is still reported by many industrial workers that verbal communication is impeded by such devices being worn in the working environment. Two experiments were carried out to investigate the effects of both earplugs and earmuffs on speech intelligibility for the listener and on the voice level of the talker in the presence of broadband masking noise. It was found that hearing protectors do not degrade speech intelligibility for the listener and may even effect a slight improvement. However, if worn by the talker, his speech is reduced in level and quality to such an extent that intelligibility for the listener is also reduced. When hearing protectors are worn by both talker and listener, the composite effect is an overall reduction in speech intelligibility compared to the open ear condition.