Methods of Improving Speech Intelligibility for Listeners with Hearing Resolution Deficit
© Kupryjanow and Czyzewski; licensee BioMed Central Ltd. 2012
Received: 25 July 2012
Accepted: 23 September 2012
Published: 25 September 2012
Methods developed for real-time time scale modification (TSM) of speech signal are presented. They are based on the non-uniform, speech rate depended SOLA algorithm (Synchronous Overlap and Add). Influence of the proposed method on the intelligibility of speech was investigated for two separate groups of listeners, i.e. hearing impaired children and elderly listeners. It was shown that for the speech with average rate equal to or higher than 6.48 vowels/s, all of the proposed methods have statistically significant impact on the improvement of speech intelligibility for hearing impaired children with reduced hearing resolution and one of the proposed methods significantly improves comprehension of speech in the group of elderly listeners with reduced hearing resolution.
Time scale modification algorithms have been widely used for supporting various types of speech perception disorders. The main fields of TSM application are: Language Learning Impairment (LLI) [1–3], Second-Language Learning [4, 5], Central Auditory Processing Disorders (CAPD) [6–9], verbal apraxia  and aphasia . Despite the large number of works that are devoted to the influence of the time-expanded speech on several disorders, there are still deficiencies in this area. For example verbal apraxia and aphasia were only mentioned by Coyle  and Nejime  but evaluation of the TSM methods with such a group of subjects was not performed. Serve-to-profoundly hearing impaired children were examined by Uchanski et al. , but in that research, analysis of the hearing resolution impairment caused by the CAPD was not investigated.
In this work, three methods for real-time TSM, dedicated for listeners with hearing resolution deficit are presented. Effectiveness of these methods was examined for two different age groups of listeners: the hearing impaired children and the elderly persons with presbycusis. The latter group was chosen based on the assumption that main difficulties in speech comprehension in elderly people are associated with central auditory processing aspects of hearing [12, 13]. The same hypothesis was the principle of speech modification methods proposed by Nakamura  and Nejime . The former was selected because of the lack of research results in the area of analysis of relationship between the CAPD and time-expansion of speech. Uchanski  has shown that there is no statistically important impact on the speech recognition rate for the group of hearing impaired children, but in that research, hearing resolution deficit of speech perception (related to the CAPD) was not investigated. Therefore, in this paper hearing resolution of the listener was analysed in accordance with the time-expanded speech intelligibility in order to investigate the relationship between the hearing resolution deficit and speech intelligibility. We have made a hypothesis that for listeners with a reduced hearing resolution, time expansion of a fast rate speech using the proposed real-time TSM methods significantly improves speech perception.
The proposed TSM methods were designed in such a way that they could work in real-time on mobile devices. Applications of these methods may be the same as those proposed by Nakamura  (i.e. stretching the speech during television news), or by Nejime  (i.e. a portable device with built-in microphone and binaural headphones). Moreover, we propose a new application of this method, i.e. stretching the speech during the phone call (an algorithm implemented on a mobile phone). We used a smartphone as a mobile platform for implementation of the proposed methods. Tests of the capabilities of the mobile phone implementation were performed and the performance results were described in earlier papers [14–16]. In this work, influence of the proposed methods on the comprehension of speech was investigated.
The outline of the paper is as follows. In Section II, the proposed TSM methods are described. In Section III, usability of these methods is investigated and the results of speech comprehension tests for both listener groups are presented. The obtained results are discussed in Section IV and concluded in the last part of the paper.
The proposed TSM methods were designed in order to modify, in real-time, a speech signal captured by the microphone located near the speaker’s mouth or the speech signal sent from a device (e.g. a cellphone, TV etc.). Three different TSM methods for the real-time speech stretching are proposed: an uniform real-time TSM (algorithm A) described by authors in an earlier paper , and two non-uniform real-time TSMs. One of the non-uniform TSMs was described in a conference paper (algorithm B) , while the second method provides a novel solution (algorithm C).
All the proposed TSM methods are based on the assumption that the input signal contains redundant information, i.e. silence passages (pauses between words, sentences, speeches) and prolonged vowels. These parts of the signal may be removed or at least they should not be stretched. This approach allows saving extra time in which the stretched speech could be presented.
where Sa is the time shift of the frame used during the analysis step, Ss is the time shift of the frame used during the synthesis step. If the value of α(t) is greater than 1, the input signal will be stretched, if α(t) is lower than 1, the signal will be shortened; for α(t) equal to 1, the time scale modification will not be performed. Since the TSM will be performed only in order to expand the time of the input signal, α(t) will take values equal or higher than 1.
Uniform speech stretching (method A)
In this method, a speech signal is stretched using constant values of the scaling factor. Input signal is time-extended only when the voice is detected by the VAD and vowel prolongation was not observed by the vowels detector. Despite the fact that the input signal is non-uniformly time scaled (silence and speech passages are modified using different stretching factor values), the speech signal is modified uniformly (with a single value of the stretching factor). The stretching procedure is controlled by the αd parameter (representing desired scaling factor). The value of αd should be specified (in the experiment it was set to a constant value equal to 1.5). Additionally, elimination of redundancy in the input signal is performed by replacing intervals of silence longer than 200 ms with the time-expanded speech.
Non-uniform TSM controlled by a scaling factor (method B)
Values of the scaling factor used in method B
Non-uniform TSM controlled by estimated ROS (method C)
where αcons(t) is the value of scaling factor for the current frame (provided a consonant was detected), αvowel(t) is the value of scaling factor for the current frame (provided a vowel was detected) , Δt is the time interval used for the ROS estimation (in the experiment, it was set to 1.5 s), Δtvowel is the duration of the vowel in the estimation interval, η is the ratio between the scaling factor used for the vowels and the scaling factor used for consonants (in the experiment, it was equal to 1.7).
An analysis of Figure 2 shows that the lowest difference in the duration of the stretched and the original speech is obtained using the method B. The method A produces the highest differences in the utterance duration. If not much redundancy is found in the input signal (using detectors), the signal can be time-expanded for a relatively long time and differences between the input and the output signal can drift towards infinity. To prevent such a situation, the TSM procedure is turned off after the difference between the input signal and output signal is higher than Δtoff, and the unmodified speech is send to the output. This threshold is exceeded much often for the method A than for methods B and C and its value can be defined by the user. During the experiments, Δtoff was set to 3 seconds.
Evaluation of the proposed methods of TSM was performed employing the sentence intelligibility test (SIT). A word recognition test was not performed, because as it was shown by the Nejime , time expansion of speech has no impact on the intelligibility of separated words.
Speech intelligibility test (SIT)
In case of SIT, 4 different types of speech were examined, i.e. the original speech and the speech stretched using three proposed TSM methods. As the speech material, Polish matrix test (PMT)  for elderly listeners and pediatric Polish matrix test (PPMT) for children were used. Usability of these tests for speech intelligibility measurements was examined and proved by the authors of the mentioned tests . In both matrix tests, each sentence has the same grammatical structure. Sentences consist of 5 words for PMT and 3 words for PPMT. The procedure of sentence creation is the same as for the typical matrix test designed by Hagerman : the list of words is fixed (to 50 words for PMT, and to 48 words for PPMT) and sentences are created by a random selection of words according to the sentence structure. This approach produced 100000 different sentences for PMT and 256 different sentences for PPMT (for more details, see papers [20, 21, 23]). The words necessary for both tests were recorded in a voice recording studio by a male speaker. All sets of words were recorded in three different average rates of speech measured in vowels/s, namely: 2.72, 4.88, 6.48 for PMT and 3.56, 6.43 and 7.58 for PPMT. Two highest rates for PPMT (ROSmean1 = 7.58 vowels/s, ROSmean2 = 6.43 vowels/s) and PMT (ROSmean3 = 6.48 vowels/s, ROSmean4 = 4.88 vowels/s) were used as the input signal during the experiment. In case of SIT, sentences were divided into two separate sets. The first set contained 40 sentences spoken with the highest speech rate (10 sentences for one type of algorithm: original, A, B, C), and the second set included 40 sentences with the second highest speech rate. During the test, each listener had to repeat words constituting the sentences. The word error rate (WER), as well as average improvement in speech intelligibility were measured. For WER calculation, the percentage of words repeated incorrectly was used, while the improvement of speech intelligibility, obtained for the proposed methods, was calculated as a difference between the WERs for the original and for the time-expanded speech.
Time compressed speech test (TCST)
Additionally, each listener performed a time-compressed speech test (TCST) in order to obtain their individual 50% time-compressed speech threshold (TCT50) defined by Versfeld , as the alternative of SRT50 (Speech Reception Threshold) . Speech material in this test was the same as in the SIT test. Since the rate of speech is artificially increased during the TCST, the average ROS of the input speech should be as low as possible to ensure a wide range of ROS values. Therefore, the average ROS of the input speech used for the test was equal to 3.56 vowels/s for the PPMT and 2.72 vowels/s for the PMT. It was observed that the chosen values of speech rate are perceived as a slow one. Originally, the TCT50 value represented a threshold, defined in syllables/s, for which 50% of the sentences in the test were correctly recognized by the listener. In our research, we have used a speech rate expressed as number of vowels/s. Thus, the speech rate defined by us is the derivation of the number of syllables/s.
The main difference between the test proposed by Versfeld  and a standard time-compressed speech test  is that in the standard test, its output provides the value of stretching factor which is independent from the rate of the input speech. Consequently, the results of this test for different speech materials cannot be compared with each other and they do not provide information about the ROS which is suitable for the listener. In turn, the results of the test proposed by Versfeld provide this kind of information, so it can be directly linked with intelligibility of time-expanded speech. The procedure related to the TCST test proposed by the Versfeld is as follows:
each listener has to repeat 13 sentences
the scaling factor is modified for each sentence according to the rules:
if all words in the last sentence were repeated correctly, the value of scaling factor increases,
otherwise the value of scaling factor decreases.
TCT50 is calculated as a geometric average of the last 10 average rates of the sentences.
Groups of listeners
For the elderly listeners, a tonal audiometry was performed in order to obtain their hearing level threshold. Listeners were using binaural headphones during the TCST and SIT and the signal level was set to a comfortable value. Each listener was asked at the beginning of test if the speech level is appropriate for them and if not, the level was adjusted. All listeners with hearing aids (HA) and cochlear implants (CI) were using their devices during the experiment.
The group of elderly people was examined during one session which lasted about 40 minutes. Tests for children were conducted in two separate sessions, because it was difficult for them to concentrate on the test for so long. Examination of children was divided into part one: TCST and SIT performed for the input speech with the average ROS equal to ROSmean1 (this part lasted about 20 minutes); and part two: SIT performed for the input speech with the average ROS equal to ROSmean2 (this part lasted about 12 minutes). The hearing thresholds were provided by the audiologist who performed this examination earlier.
In both groups of listeners, 17 volunteers were investigated. The average age of hearing impaired children was 9.35 years (9 females, 8 males), whereas the average age of elderly listeners was 80.76 years (11 females, 6 males).
TCST results obtained for the hearing impaired children and the elderly listeners
95% confidence interval
Hearing impaired children
Analysis of the SIT was performed separately for the group of the hearing impaired children and the elderly listeners. Statistical importance of the differences between mean values of the WER obtained for the original speech and the modified one was examined using an one-way repeated measures ANOVA (RM ANOVA) test and the Friedman’s test. For all analyses, a normal distribution of data was checked using the Shapiro-Wilk test and the hypothesis of the sphericity was verified using the Mauchly's test. The RM ANOVA test was performed only when both assumptions (normal distribution and sphericity) were met, otherwise the non-parametric Friedman test was used. For all tests, the significance level equal to 0.05 was assumed.
Hearing impaired children
Results of hearing tests obtained for the hearing impaired children with reduced hearing resolution threshold (TCT 50 < 5.71 vowels/s)
Hearing thresholds [dB HL]
WER [%] ROSmean1
WER [%] ROSmean2
Results of hearing tests obtained for the hearing impaired children with normal hearing resolution threshold (TCT 50 ≥ 5.71 vowels/s)
Hearing thresholds [dB HL]
WER [%] ROSmean1
WER [%] ROSmean2
It can be seen that WER obtained for the speech spoken with the average ROS equal to ROSmean2 is much lower than for the average ROSmean1. This relation is valid for both subgroups of the hearing impaired children (the normal and the reduced hearing resolution). It may be related to the fact that the ROSmean2 value (6.43 vowels/s) is close to the average values of the TCT50 obtained for both subgroups of the hearing impaired children (the subgroup with reduced hearing resolution μ(TCT50) = 5.16 vowels/s and the subgroup with normal μ(TCT50) = 6.98 vowels/s). Furthermore, the ROSmean1 value (7.48 vowels/s) is higher than the average values of the TCT50 achieved in both subgroups of children. It should be also pointed out that the average WER obtained in two subgroups for the speech spoken with the ROS equal to ROSmean1 is similarly high (56.11% for TCT50 < 5.71; 40.15% for TCT50 ≥ 5.71) and for the ROS equal to ROSmean2, it is comparatively small (17.78%, 13.63%, respectively). Based on these observations, the following conclusion can be made: the hearing impaired children in both subgroups had comparable problems with comprehension of the unmodified speech.
A significant improvement in the intelligibility can be mostly seen for the input speech spoken with average ROS equals to ROSmean1 (Figure 5) and speech modification algorithms A and C (from 6.43% for the children with normal hearing resolution and algorithm C to 27.63% for the algorithm A and children with the reduced hearing resolution). For the speech modified using the algorithm B, only the subgroup of children with reduced hearing resolution shows high improvement in speech intelligibility (27.77%).
To verify if the differences of the average WER values are statistically important, appropriate analyses were performed (separately for the input speech of ROSmean1 and ROSmean2). For the input ROS equal to ROSmean1, the results of SIT obtained by the subgroup of children with TCT50 < 5.71 vowels/s did not show the normal distribution. Therefore, for these data, Friedman’s non-parametric test was applied. The statistic value of the test was equal to χ2(3) = 113.5. Since the number of listeners in this subgroup was low (6 persons), in order to increase the reliability of the result, the p-value was not calculated but the test statistic value was compared with the suitable critical value read from the tables (χ2(3)cv = 76). Hence, the critical test value is lower than the obtained Friedman’s statistic value, the differences between the WERs obtained for the unmodified speech and the speech modified using algorithm A-C are statistically important. In order to check which algorithm causes this situation, a post hoc non-parametric test equivalent to the Least Significant Difference Fisher’s test was performed. For the pairs of unmodified speech and speech modified using algorithms from A to C, the following statistical values were obtained: 12.5 (A), 13.5 (B), and 8.0 (C). These results were compared with the critical value equal to 6.2. Since all the statistics values were higher than the critical value for all the proposed algorithms, differences between the WER obtained for the unmodified and modified speech are statistically important.
For the same speech rate and the subgroup of children with TCT50 ≥ 5.71 vowels/s, the normal distribution was confirmed (for all algorithms) and the assumption of sphericity was met. For these reasons, RM ANOVA was calculated for these data. The following results were obtained: F(3,30) = 3.44 and p = 0.03. Since the achieved p value is lower than the assumed significance level in at least one pair, the differences between average values of WER are statistically important. Additionally, the post hoc LSD test shows that only differences in average WER between the pairs ‘original-method A’ (t(10) = −2.83; p = 0.02) and ‘method A- method B’ (t(10) = 2.54); p = 0.03) are statistically important. This results shows that for the subgroup of hearing impaired children with the normal hearing resolution, method A gives statistically important improvement of in speech intelligibility.
In case of the speech spoken with ROSmean2, improvement in speech ineligibility was observed only for the subgroup of children with reduced hearing resolution and methods A-B (3.89% and 3.34%, respectively) and for the subgroup of children with normal hearing resolution and method C (3.94%). In other cases, a slight decrease in WER was observed (from −5.55% to −0.3%). For the subgroup of children with reduced hearing resolution, RM ANOVA was calculated (data had normal distribution and assumption of sphericity was met). The results of the test show that there are no statistically important differences in WER between the analysed methods (F(3,15) = 1.51; p = 0.25). For the subgroup of children with normal hearing resolution, the Friedman’s test was performed (data were not normally distributed) and there was no statistically important difference in WER between the methods (χ2(3) = 0.30; p = 0.82). Therefore, none of the proposed methods affect the intelligibility of speech spoken with ROS equal to ROSmean2.
Results of hearing tests obtained for the elderly listeners with normal hearing resolution threshold (TCT 50 < 3.99 vowels/s)
Mean audiometric hearing thresholds [dB HL]
WER [%] ROSmean3
WER [%] ROSmean4
Results of hearing tests obtained for the elderly listeners with normal hearing resolution threshold (TCT 50 ≥ 3.99 vowels/s)
Mean audiometric hearing thresholds [dB HL]
WER [%] ROSmean3
WER [%] ROSmean4
Since the observed improvement in speech comprehension was statistically important only for the method B, a relationship between the TCT50 and the intelligibility improvement was analysed only for these data. In Figure 12, this relationship is presented. The calculated Pearson correlation coefficient was equal to 0.58 and its positive value indicates that the improvement in speech comprehension (provided by the method B) is higher when the hearing resolution of listener increases. This observation is surprising because the inverse relationship was expected (the higher hearing resolution, the lower improvement of speech comprehension). This phenomenon may be caused by the fact that in the group of elderly listeners with reduced hearing resolution, only one subject was using HA and the hearing losses of all the listeners in this subgroup were also significant (see Table 5). Consequently, two hearing impairments overlap here and cause the difficulties in speech comprehension.
The speech intelligibility test performed in two groups of listeners (the hearing impaired children and the elderly listeners) have shown that there are differences in speech comprehension in case of time scale modified speech in comparison to the original one. These differences are significant only if a very fast speech is used as the input signal for the test (ROSmean1, ROSmean3). For the hearing impaired children with reduced hearing resolution, all the proposed methods gave statistically important improvement. For the group of elderly listeners, only the method B produces statistically important improvement of speech comprehension. The importance of this improvement was proved in the group of listeners with very low hearing resolution (TCT50 < 3.99 vowels/s). These results are similar to those presented by Nejime et al. , who verified their method for listeners with reduced hearing resolution (measured using RGDT - Random Gap Detection Test). Nejime et al. showed the importance of speech modification using their method, which stretches only the high energy parts of speech. Nejime did not found a clear relationship between the hearing impairment or hearing resolution and the results of WER. This relationship was shown in our research. First, for the subgroup of hearing impaired children with reduced hearing resolution, a significant correlation between TCT50 and WER was observed. Second, a not so high but significant correlation was found for the relation between TCT50 and the improvement (in subgroups with deficit in hearing resolution) of intelligibility of speech stretched using the method B for elderly listeners and the method A for the hearing impaired children.
For the speech spoken with lower rates (ROSmean2, ROSmean4), the obtained results are consistent with those presented by Uchanski et al. . There are no statistically important differences in speech intelligibility between the original speech and the time-expanded one.
Three methods for real-time speech stretching were proposed and verified experimentally. It was proved that the method B significantly improves speech comprehension in hearing impaired children, as well as in elderly people with deficit in the hearing resolution. In turn, the proposed non-uniform real-time speech stretching method A brings satisfying results not only for the hearing impaired children with low value of TCT50, but also for children with normal hearing resolution. The presented results are in a good accordance with the state-of-the-art results and extend them with the introduction of analysis of the impact of the input speech rate on the relation between the TSM and speech intelligibility. Another novelty is the usage of TCT50 as a measure of the hearing resolution deficit. It was shown that this parameter correlates with the improvement achieved by employing time-expanded speech.
Time Scale Modification
Language Learning Impaired
Central Auditory Processing Disorders
Voice Activity Detection
Synchronous Overlap and Add
Rate of Speech
Sentence Intelligibility Test
Polish Matrix Test
Pediatric Polish Matrix Test
Word Error Rate
Time-Compressed Speech Test
Speech Reception Threshold
50% time-compressed speech threshold
Analysis Of Variance
ANOVA: Repeated Measures ANOVA
Least Significant Difference
Random Gap Detection Test.
The research was funded within the project No. POIG.01.03.01-22-017/08, entitled "Elaboration of a series of multimodal interfaces and their implementation to educational, medical, security and industrial applications". The project is subsidized by the European Regional Development Fund by the Polish State budget".
- Nagarajan SS, Wang X, Merzenich MM, Schreiner CE, Johnston P, Jenkins WM, Miller S, Tallal P: Speech modifications algorithms used for training language learning-impaired children. Ieee Transactions On Rehabilitation Engineering. 1998, 6: 257-268. 10.1109/86.712220.View ArticlePubMedGoogle Scholar
- Tallal P, Miller SL, Bedi G, Byma G, Wang X, Nagarajan SS, Schreiner C, Jenkins WM, Merzenich MM: Language comprehension in language-learning impaired children improved with acoustically modified speech. Science. 1996, 271: 81-84. 10.1126/science.271.5245.81.View ArticlePubMedGoogle Scholar
- Eroguli O, Karagoz I: Time-scale modification of speech signals for language-learning impaired children. 1998, Istanbul: IEEE: In Proceedings of the Biomedical Engineering Days 2nd International ConferenceView ArticleGoogle Scholar
- Donnellan O, Jung E, Coyle E: Speech-adaptive time-scale modification for computer assisted language-learning. Proceedings of the Third IEEE International Conference on Advanced Learning Technologies (ICALT'03). 2003Google Scholar
- Yang H, Guoi W, Liang Q: A speaking rate adjustable digital speech repeater for listening comprehension in second-language learning. 2008, Wuhan: IEEE Computer Society Washington: In Proceedings of the International Conference on Computer Science and Software EngineeringView ArticleGoogle Scholar
- Nakamura A, Seiyama N, Takagi T, Miyasaka E: Real time speech rate converting system for elderly people. Proceedings of the IEEE Int. Conf. Acoust., Speech, Signal Proc (ICASSP). 1994, AdelaideGoogle Scholar
- Nejime Y, Aritsuka T, Imamura T, Hokimoto T, Ifukubei T, Matsushima J: A portable digital speech-rate converter for elderly hearing-impaired listeners. Proceeding of 16th Annual International Conference of the IEEE. 1994Google Scholar
- Nejime Y, Moore BCJ: Evaluation of the effect of speech-rate slowing on speech intelligibility in noise using a simulation of cochlear hearing loss. J Acoust Soc Am. 1998, 103: 572-576. 10.1121/1.421123.View ArticlePubMedGoogle Scholar
- Nejime Y, Aritsuka T, Imamura T, Ifukubei T, Matsushima J: A portable digital speech-rate converter for hearing impairment. IEEE Trans. Rehabil. Eng. 1996, 4: 73-83. 10.1109/86.506404.View ArticlePubMedGoogle Scholar
- Coyle E, Donnellan O, Jung E, Meinardi M, Campbell D, MacDonailli C, Leung PK: Time-scale modification as a speech therapy tool for children with verbal apraxia. Proceedings of the 5th Intl. Conf. Disability, Virtual Reality and Assoc. Tech. 2004, OxfordGoogle Scholar
- Uchanski RM, Geersi AE, Protopapas A: Intelligibility of modified speech for young listeners with normal and impaired hearing. Journal of Speech, Language, and Hearing Research. 2002, 45: 1027-1038. 10.1044/1092-4388(2002/083).View ArticlePubMedGoogle Scholar
- Rooji JCGM, Promp R: Auditive and cognitive factors in speech perception by elderly listeners. I1:Multivariate analysis. J Acoust Soc Am. 1990, 88: 2611-2624. 10.1121/1.399981.View ArticleGoogle Scholar
- Satoh T, Adachi T: Identification of time required by elderly individual for the perception of vowels. Audiology Japan. 1988, 31: 737-743. 10.4295/audiology.31.737.View ArticleGoogle Scholar
- Kupryjanow A, Czyzewski A: A non-uniform real-time speech time-scale stretching method. Proceedings of International Conference on Signal Processing and Multimedia Applications (SIGMAP). 2011, SevilleGoogle Scholar
- Kupryjanow A, Czyzewski A: Real-time speech-rate modification experiments. Audio Engineering Society Convention Paper, preprint No. 8052. 2010, LondonGoogle Scholar
- Kupryjanow A, Czyzewski A: Improved method for real-time speech stretching”. Intelligent Decision Technologies. 2012, 6: 177-185.Google Scholar
- Chu WC, Lashkari K: Energy-based nonuniform time-scale compression of audio signals. IEEE Trans Consum Electron. 2003, 49: 183-187. 10.1109/TCE.2003.1205474.View ArticleGoogle Scholar
- Demol M, Verhelst W, Struyei K, Verhoeve P: Efficient non-uniform time-scaling of speech with WSOLA. Proceedings of the Speech and Computers (SPECOM). 2005Google Scholar
- Dorran D, Lawlor R, Coyle E: A comparison of time-domain time-scale modification algorithms. Audio Engineering Society Convention Paper, preprint No. 6674. 2006, ParisGoogle Scholar
- Ozimek E, Kutzner D, Libiszewski P, Warzyboki A, Kociński J: The new polish tests fo speech intelligibility measurements. Int J Audiol. 2010, 49: 444-454. 10.3109/14992021003681030.View ArticlePubMedGoogle Scholar
- Ozimek E, Libiszewski P, Kutzner D: Polski pediatryczny test zdaniowy do pomiarów zrozumiałości mowy prezentowanej na tle szumu. Biuletyn Polskiego Stowarzyszenia Protetyków Słuchu. 2010, 40: 9-13.Google Scholar
- Hagerman B: Sentences for testing speech intelligibility in noise. Scand. Sudiol. 1982, 11: 79-87.View ArticleGoogle Scholar
- Plompi R, Mimpen AM: Improving the reliability of testing the speech reception threshold for sentences. Audiology. 1979, 18: 43-52. 10.3109/00206097909072618.View ArticleGoogle Scholar
- Versfeld NJ, Dreschler WA: The relationship between the intelligibility of time-compressed speech and speech in noise in young and elderly listeners. J Acoust Soc Am. 2002, 111: 401-408. 10.1121/1.1426376.View ArticlePubMedGoogle Scholar
- Keith RW: Standardization of the time compressed sentence test. Journal of Educational Audiology. 2002, 10: 15-20.Google Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.