汉语语音合成系统评价方法

从1994年开始,对汉语语音合成系统的工作性能定期举行全国评测.采用语言清晰度测试方法,1994年对五个不同的合成系统进行了评测和诊断.听音人为16名大学生(男8,女8),对合成言语没有经验.听音人响应是开放的听音记录.同时,还采用十点主观评价(MOS)测定言语自然度.为给出各合成系统音段层的诊断信息,对合成语音的辅音知觉混淆矩阵进行了分析.借助于对比自然言语和合成言语在不同语言层次上清晰度试验得分间的统计关系,来考察合成系统韵律特征处理的缺陷.结果表明,采用上述方法可得到评测合成系统工作性能的稳定合理的指标.有关韵律特征的评价方法有待于进一步发展.     

Speech signal modification to increase intelligibility in noisy environments

The role of transient speech components on speech intelligibility was investigated. Speech was decomposed into two components--quasi-steady-state (QSS) and transient--using a set of time-varying filters whose center frequencies and bandwidths were controlled to identify the strongest formant components in speech. The relative energy and intelligibility of the QSS and transient components were compared to original speech. Most of the speech energy was in the QSS component, but this component had low intelligibility. The transient component had much lower energy but was almost as intelligible as the original speech, suggesting that the transient component included speech elements important to speech perception. A modified version of speech was produced by amplifying the transient component and recombining it with the original speech. The intelligibility of the modified speech in background noise was compared to that of the original speech, using a psychoacoustic procedure based on the modified rhyme protocol. Word recognition rates for the modified speech were significantly higher at low signal-to-noise ratios (SNRs), with minimal effect on intelligibility at higher SNRs. These results suggest that amplification of transient information may improve the intelligibility of speech in noise and that this improvement is more effective in severe noise conditions.     

Estimating speech spectra for copy synthesis by linear prediction and by hand

Linear prediction is a widely available technique for analyzing acoustic properties of speech, although this method is known to be error-prone. New tests assessed the adequacy of linear prediction estimates by using this method to derive synthesis parameters and testing the intelligibility of the synthetic speech that results. Matched sets of sine-wave sentences were created, one set using uncorrected linear prediction estimates of natural sentences, the other using estimates made by hand. Phoneme restrictions imposed on linguistic properties allowed comparisons between continuous and intermittent voicing, oral or nasal and fricative manner, and unrestricted phonemic variation. Intelligibility tests revealed uniformly good performance with sentences created by hand-estimation and a minimal decrease in intelligibility with estimation by linear prediction due to manner variation with continuous voicing. Poorer performance was observed when linear prediction estimates were used to produce synthetic versions of phonemically unrestricted sentences, but no similar decline was observed with synthetic sentences produced by hand estimation. The results show a substantial intelligibility cost of reliance on uncorrected linear prediction estimates when phonemic variation approaches natural incidence.     

Speech intelligibility of English,Polish, Arabic and Mandarin under different room acoustic conditions

This paper examines the impact of room acoustic conditions on the speech intelligibility of four languages (English, Polish, Arabic and Mandarin). Listening test scores (diagnostic rhyme tests, phonemically balanced word tests and phonemically balanced sentence tests) of the four languages were compared under four room acoustic conditions defined by their speech transmission index (STI = 0.2, 0.4, 0.6 and 0.8). The results obtained indicated that there was a statistically significant difference between the word intelligibility scores of languages under all room acoustic conditions, apart from the STI = 0.8 condition. English was the most intelligible language under all conditions, and differences with other languages were larger when conditions were poor (maximum difference of 29% at STI = 0.2, 33% at STI = 0.4 and 14% at STI = 0.6). Results also showed that Arabic and Polish were particularly sensitive to background noise, and that Mandarin was significantly more intelligible than those languages at STI = 0.4. Consonant-to-vowel ratios and languages’ distinctive features and acoustical properties explained some of the scores obtained. Sentence intelligibility scores confirmed variations between languages, but these variations were statistically significant only at the STI = 0.4 condition (sentence tests being less sensitive to very good and very poor room acoustic conditions). Overall, the results indicate that large variations between the speech intelligibility of different languages can occur, especially for spaces that are expected to be challenging in terms of room acoustic conditions. Recommendations solely based on room acoustic parameters (e.g. STI) might then prove to be insufficient for designing a multilingual environment.     

Bidirectional clear speech perception benefit for native and high-proficiency non-native talkers and listeners: intelligibility and accentedness

This study investigated how native language background interacts with speaking style adaptations in determining levels of speech intelligibility. The aim was to explore whether native and high proficiency non-native listeners benefit similarly from native and non-native clear speech adjustments. The sentence-in-noise perception results revealed that fluent non-native listeners gained a large clear speech benefit from native clear speech modifications. Furthermore, proficient non-native talkers in this study implemented conversational-to-clear speaking style modifications in their second language (L2) that resulted in significant intelligibility gain for both native and non-native listeners. The results of the accentedness ratings obtained for native and non-native conversational and clear speech sentences showed that while intelligibility was improved, the presence of foreign accent remained constant in both speaking styles. This suggests that objective intelligibility and subjective accentedness are two independent dimensions of non-native speech. Overall, these results provide strong evidence that greater experience in L2 processing leads to improved intelligibility in both production and perception domains. These results also demonstrated that speaking style adaptations along with less signal distortion can contribute significantly towards successful native and non-native interactions.     

Performance of hearing-impaired listeners under various types of amplitude compression

Speech perception by subjects with sensorineural hearing impairment was studied using various types of short-term (syllabic) amplitude compression. Average speech level was approximately constant. In quiet, a single-channel wideband compression (WBC) with compression ratio equal to 10, attack time 10 ms and release time 90 ms produced significantly higher scores than a three-channel multiband compression (MBC) or no compression when a nonsense syllable test (City University of New York) was used. The scores under MBC, WBC, or no compression were not significantly different when the modified rhyme test (MRT) was used. But when overshoots caused by compression were clipped, the MRT scores improved significantly. The influence of MBC on reverberant speech and of WBC on noisy speech were tested with the MRT. Reverberation reduced the scores, and this reduction was the same with compression as without. Noise added to speech before compression also reduced the scores, but the reduction was larger with compression than without. When noise was added after compression, an improvement was observed when WBC had a compression ratio of about 5, attack time 1 ms, and release time 30 ms. Other compression modes (e.g., with high-frequency pre-emphasis) did not show an improvement. The results indicate that WBC with a compression ratio around 5, attack time shorter than 3 ms, and release time between 30 and 90 ms can be beneficial if signal-to-noise ratio is large, or, if in a noisy or reverberant environment, the effects of noise or reverberation are eliminated by using listening systems.     

A method to determine the speech transmission index from speech waveforms

A method for computing the speech transmission index (STI) using real speech stimuli is presented and evaluated. The method reduces the effects of some of the artifacts that can be encountered when speech waveforms are used as probe stimuli. Speech-based STIs are computed for conversational and clearly articulated speech in several noisy, reverberant, and noisy-reverberant environments and compared with speech intelligibility scores. The results indicate that, for each speaking style, the speech-based STI values are monotonically related to intelligibility scores for the degraded speech conditions tested. Therefore, the STI can be computed using speech probe waveforms and the values of the resulting indices are as good predictors of intelligibility scores as those derived from MTFs by theoretical methods.     

The performance of different synthesis signals in acoustic models of cochlear implants

Synthesis (carrier) signals in acoustic models embody assumptions about perception of auditory electric stimulation. This study compared speech intelligibility of consonants and vowels processed through a set of nine acoustic models that used Spectral Peak (SPEAK) and Advanced Combination Encoder (ACE)-like speech processing, using synthesis signals which were representative of signals used previously in acoustic models as well as two new ones. Performance of the synthesis signals was determined in terms of correspondence with cochlear implant (CI) listener results for 12 attributes of phoneme perception (consonant and vowel recognition; F1, F2, and duration information transmission for vowels; voicing, manner, place of articulation, affrication, burst, nasality, and amplitude envelope information transmission for consonants) using four measures of performance. Modulated synthesis signals produced the best correspondence with CI consonant intelligibility, while sinusoids, narrow noise bands, and varying noise bands produced the best correspondence with CI vowel intelligibility. The signals that performed best overall (in terms of correspondence with both vowel and consonant attributes) were modulated and unmodulated noise bands of varying bandwidth that corresponded to a linearly varying excitation width of 0.4 mm at the apical to 8 mm at the basal channels.     

Recognition of accented English in quiet and noise by younger and older listeners

This study investigated the effects of age and hearing loss on perception of accented speech presented in quiet and noise. The relative importance of alterations in phonetic segments vs. temporal patterns in a carrier phrase with accented speech also was examined. English sentences recorded by a native English speaker and a native Spanish speaker, together with hybrid sentences that varied the native language of the speaker of the carrier phrase and the final target word of the sentence were presented to younger and older listeners with normal hearing and older listeners with hearing loss in quiet and noise. Effects of age and hearing loss were observed in both listening environments, but varied with speaker accent. All groups exhibited lower recognition performance for the final target word spoken by the accented speaker compared to that spoken by the native speaker, indicating that alterations in segmental cues due to accent play a prominent role in intelligibility. Effects of the carrier phrase were minimal. The findings indicate that recognition of accented speech, especially in noise, is a particularly challenging communication task for older people.     

Improvement of intelligibility of ideal binary-masked noisy speech by adding background noise

When a target-speech/masker mixture is processed with the signal-separation technique, ideal binary mask (IBM), intelligibility of target speech is remarkably improved in both normal-hearing listeners and hearing-impaired listeners. Intelligibility of speech can also be improved by filling in speech gaps with un-modulated broadband noise. This study investigated whether intelligibility of target speech in the IBM-treated target-speech/masker mixture can be further improved by adding a broadband-noise background. The results of this study show that following the IBM manipulation, which remarkably released target speech from speech-spectrum noise, foreign-speech, or native-speech masking (experiment 1), adding a broadband-noise background with the signal-to-noise ratio no less than 4 dB significantly improved intelligibility of target speech when the masker was either noise (experiment 2) or speech (experiment 3). The results suggest that since adding the noise background shallows the areas of silence in the time-frequency domain of the IBM-treated target-speech/masker mixture, the abruption of transient changes in the mixture is smoothed and the perceived continuity of target-speech components becomes enhanced, leading to improved target-speech intelligibility. The findings are useful for advancing computational auditory scene analysis, hearing-aid/cochlear-implant designs, and understanding of speech perception under "cocktail-party" conditions.     

Traffic noise annoyance and speech intelligibility in persons with normal and persons with impaired hearing

Annoyance ratings in speech intelligibility tests at 45 dB(A) and 55 dB(A) traffic noise were investigated in a laboratory study. Subjects were chosen according to their hearing acuity to be representative of 70-year-old men and women, and of noise-induced hearing losses typical for a great number of industrial workers. These groups were compared with normal hearing subjects of the same sex and, when possible, the same age. The subjects rated their annoyance on an open 100 mm scale. Significant correlations were found between annoyance expressed in millimetres and speech intelligibility in percent when all subjects were taken as one sample. Speech intelligibility was also calculated from physical measurements of speech and noise by using the articulation index method. Observed and calculated speech intelligibility scores are compared and discussed. Also treated is the estimation of annoyance by traffic noise at moderate noise levels via speech intelligibility scores.     

Perceptual contributions to monosyllabic word intelligibility: segmental, lexical, and noise replacement factors

This study investigated the relative contributions of consonants and vowels to the perceptual intelligibility of monosyllabic consonant-vowel-consonant (CVC) words. A noise replacement paradigm presented CVCs with only consonants or only vowels preserved. Results demonstrated no difference between overall word accuracy in these conditions; however, different error patterns were observed. A significant effect of lexical difficulty was demonstrated for both types of replacement, whereas the noise level used during replacement did not influence results. The contribution of consonant and vowel transitional information present at the consonant-vowel boundary was also explored. The proportion of speech presented, regardless of the segmental condition, overwhelmingly predicted performance. Comparisons were made with previous segment replacement results using sentences [Fogerty, and Kewley-Port (2009). J. Acoust. Soc. Am. 126, 847-857]. Results demonstrated that consonants contribute to intelligibility equally in both isolated CVC words and sentences. However, vowel contributions were mediated by context, with greater contributions to intelligibility in sentence contexts. Therefore, it appears that vowels in sentences carry unique speech cues that greatly facilitate intelligibility which are not informative and/or present during isolated word contexts. Consonants appear to provide speech cues that are equally available and informative during sentence and isolated word presentations.     

Application of an extended equalization-cancellation model to speech intelligibility with spatially distributed maskers

An extended version of the equalization-cancellation (EC) model of binaural processing is described and applied to speech intelligibility tasks in the presence of multiple maskers. The model incorporates time-varying jitters, both in time and amplitude, and implements the equalization and cancellation operations in each frequency band independently. The model is consistent with the original EC model in predicting tone-detection performance for a large set of configurations. When the model is applied to speech, the speech intelligibility index is used to predict speech intelligibility performance in a variety of conditions. Specific conditions addressed include different types of maskers, different numbers of maskers, and different spatial locations of maskers. Model predictions are compared with empirical measurements reported by Hawley et al. [J. Acoust. Soc. Am. 115, 833-843 (2004)] and by Marrone et al. [J. Acoust. Soc. Am. 124, 1146-1158 (2008)]. The model succeeds in predicting speech intelligibility performance when maskers are speech-shaped noise or broadband-modulated speech-shaped noise but fails when the maskers are speech or reversed speech.     

Monaural speech segregation using synthetic speech signals

When listening to natural speech, listeners are fairly adept at using cues such as pitch, vocal tract length, prosody, and level differences to extract a target speech signal from an interfering speech masker. However, little is known about the cues that listeners might use to segregate synthetic speech signals that retain the intelligibility characteristics of speech but lack many of the features that listeners normally use to segregate competing talkers. In this experiment, intelligibility was measured in a diotic listening task that required the segregation of two simultaneously presented synthetic sentences. Three types of synthetic signals were created: (1) sine-wave speech (SWS); (2) modulated noise-band speech (MNB); and (3) modulated sine-band speech (MSB). The listeners performed worse for all three types of synthetic signals than they did with natural speech signals, particularly at low signal-to-noise ratio (SNR) values. Of the three synthetic signals, the results indicate that SWS signals preserve more of the voice characteristics used for speech segregation than MNB and MSB signals. These findings have implications for cochlear implant users, who rely on signals very similar to MNB speech and thus are likely to have difficulty understanding speech in cocktail-party listening environments.     

Feasibility of subjective speech intelligibility assessment based on auralization

Subjective speech intelligibility can be assessed by speech recorded in an anechoic chamber and then convolved with room impulse responses that can be created by acoustic simulation. The speech intelligibility (SI) assessment based on auralization was validated in three rooms. The articulation scores obtained from simulated sound field were compared with the ones from measured sound field and from direct listening in rooms. Results show that the speech intelligibility prediction based on auralization technique with simulated binaural room impulse responses (BRIRs) is in agreement with reality and results from measured BRIRs. When this technique is used with simulated and measured monaural room impulse responses (MRIRs), the predicted results underestimate the reality. It has been shown that auralization technique with simulated BRIRs is capable of assessing subjective speech intelligibility of listening positions in the room.     

The effects of air temperature and humidity on the acoustic design of voice alarm systems on underground stations

Sound attenuation of air due to climatic conditions is often assumed to be constant and/or negligible in the electro acoustic design of voice alarm (VA) systems. However, air attenuation variations can be significant in large underground spaces and particularly as the frequency increases to the mid to high frequencies which are the most relevant to speech intelligibility. This investigation evaluates and quantifies the impact of the variability of the most influential climatic parameters, air temperature and relative humidity, on the performance of VA systems in underground stations. Computer simulations were employed to predict the effect of varying these climatic parameters on key performance metrics. Results demonstrated a significant increase in the values of reverberation time parameters with both temperature and humidity, at frequencies critical for speech intelligibility. Consequently the values of speech intelligibility related metrics decreased with rising temperatures and humidity. Hence, the study shows how ignoring temperature and humidity effects can lead to calculation errors in the design of VA systems. These errors could cause over-specification of the absorption required of surface materials, and the inaccurate prediction of acoustic and speech intelligibility related parameters.     

Comparison of amplification systems in an auditorium

Intelligibility of speech at two positions in a large auditorium was compared for the public address system (PA) and two assistive listening systems: Frequency modulation of radio frequencies (FM) and modulation of infrared light waves (IR). Listening groups were: normal-hearing adults, hearing-impaired, hearing aid users, elderly, and non-native. Word-identification scores were obtained with the Modified Rhyme Tests. Analysis of variance indicated that the main effects of systems, groups, and listening position were significant. Also significant were the two-way interactions. For all groups, the assistive listening systems provided better scores than the PA system. The difference between the two systems was statistically significant, but very small. It can be concluded that both listening systems provide improved speech intelligibility for various types of listeners.     

The effect of timing errors on the intelligibility of deaf children's speech.

The purpose of this study was to quantify the effect of timing errors on the intelligibility of deaf children's speech. Deviant timing patterns were corrected in the recorded speech samples of six deaf children using digital speech processing techniques. The speech waveform was modified to correct timing errors only, leaving all other aspects of the speech unchanged. The following six-stage approximation procedure was used to correct the deviant timing patterns: (1) original, unaltered utterances, (2) correction of pauses only, (3) correction of relative timing, (4) correction of absolute syllable duration, (5) correction of relative timing and pauses, and (6) correction of absolute syllable duration and pauses. Measures of speech intelligibility were obtained for the original and the computer-modified utterances. On the average, the highest intelligibility score was obtained when relative timing errors only were corrected. The correction of this type of error improved the intelligibility of both stressed and unstressed words within a phrase. Improvements in word intelligibility, which occurred when relative timing was corrected, appeared to be closely related to the number of phonemic errors present within a word. The second highest intelligibility score was obtained for the original, unaltered sentences. On the average, the intelligibility scores obtained for the other four forms of timing modification were poorer than those obtained for the original sentences. Thus, the data show that intelligibility improved, on the average, when only one type of error, relative timing, was corrected.