The main aim of the thesis is to find out possible relations between speech intelligibility and multi-lingual communication, in terms of acoustics, linguistics and perceptual factors. More specifically, the work focuses on the impact of room acoustic conditions on the speech intelligibility of four languages representative of a wide range of linguistic properties (English, Polish, Arabic and Mandarin). Additionally, perceived speech intelligibility and soundscape associated to these languages are also analysed.
Lastly, the study investigates several standards and design guidelines of spaces used for speech and their relation with multilingual intelligibility from the perspective of the
The objectives of the research are:
1. To understand relations between language specific effects and speech intelligibility, as well as relations between room acoustic properties and speech intelligibility of the four languages (Chapter 4).
a. To analyse differences in the word intelligibility of English, Polish, Arabic and Mandarin, for a range of room acoustic conditions.
b. To compare the intelligibility of distinctive features within and across languages tested with Diagnostic Rhyme Tests (i.e. English, Arabic and Mandarin), in order to obtain an insight into word intelligibility variations.
c. To analyse differences in the sentence intelligibility of English, Polish, Arabic and Mandarin, for a range of room acoustic conditions, and compare those with the differences observed for word intelligibility.
2. To analyse the effects of soundscapes on the perceived intelligibility of the four languages (Chapter 5).
a. To identify the variability of perceived speech intelligibility across different environments (an airport check-in area, a hospital reception and a café), for a range of room acoustic conditions.
b. To examine differences between perceived intelligibility and actual intelligibility.
c. To identify correlations between perceived speech intelligibility and semantic attributes characterizing communication as well as the acoustic environment, in order to ascertain the importance of multiple factors with regard to multilingual communication.
3. To review relevant standards and design guidelines of spaces used for speech and to critically analyse their relation with multilingual intelligibility from the perspective of the outcomes of the study (Chapter 6).
The findings obtained could ultimately be used to inform design guidelines and spatial design solutions of multilingual spaces, in order to minimise communication problems between end users.
In order to achieve the main objectives of this research project, three different methodological approaches have been used:
1. Word (DRT and PB-word) and sentence (PB-sentence) intelligibility test (to address objectives 1, 2, and 3).
2. Semantic differential tests (to address objectives 4, 5, and 6).
3. Calculating STIs based on the room acoustic parameters (reverberation time and signal-to-noise ratio) presented in the standards and design guidelines reviewed, and comparing suggestions with the outcomes of the study (to address objective 7).
1.4.1 Word and sentence intelligibility tests
Languages representative of a wide range of linguistic properties were selected from different language families such as the Indo-European (e.g. English, German, Polish, Spanish, and Farsi), Uralic (e.g. Turkish), Afro-Asiatic (e.g. Arabic), and Sino-Tibetan (e.g. Mandarin). Five criteria were applied for identifying the languages to be tested:
real environment depiction, consonant-to-vowel ratio, tonality, native speakers’
population, and availability of subjects. Four languages were selected following these criteria. These were English, Mandarin, Arabic, and Polish.
To assess the speech intelligibility of each language, diagnostic rhyme tests (DRT), phonemically balanced (PB) word lists and phonemically balanced sentence lists were used. DRT and PB word tests were employed to examine word intelligibility, whilst PB sentence tests were used for the analysis of sentence intelligibility. PB word tests were used for only Polish, because of the lack of DRT material in Polish.
The listening tests were conducted in one of the chambers of the acoustic laboratory of Heriot-Watt University. 3 male and 3 female listeners were selected from native speakers of each language, in order to achieve equal gender representation. The listeners of each language were selected from the same regions/countries of the speakers. Prior to the listening tests, hearing tests were carried out in the anechoic chamber of Heriot-Watt University to ensure that participants had normal hearing abilities. For DRT tests,
(by ticking), whilst for PB words and PB sentences, these had to be written down. Each listening test was repeated for four different acoustic conditions (STI = 0.2, 0.4, 0.6 and 0.8), by changing the reverberation time and signal-to-noise ratio.
A total of six statistical analysis methods were applied to the data sets in order to test several hypothesis of the current study; these methods were Intra-Class Correlation analysis, one-way Analysis of Variance (one-way ANOVA), factorial Analysis of Variance (factorial ANOVA), Spearman's RHO correlation analysis, and Principal Component Analysis (PCA). Consistency within the test participants taking part in the intelligibility tests was analysed by using the Intra-Class Correlation analysis. The difference between languages was statistically analysed by using the One-way ANOVA test. Factorial ANOVA was used in order to analyse the combined effects of languages and room acoustic conditions. Spearman's RHO correlation analysis was performed to investigate correlations between the consonant-to-vowel ratio of languages and word/sentence intelligibility scores. Finally, the interaction between the distinctive feature intelligibility scores (language specific word intelligibility scores) were investigated by using the PCA.
The results obtained were used to investigate the impact of room acoustic conditions on the speech intelligibility, as well as the relationship between the room acoustic parameters and distinctive features of four languages (English, Polish, Arabic and Mandarin). Additionally, the results allowed identifying the correlation between consonant-to-vowel ratios and speech intelligibility, and the relationship between word and sentence intelligibility.
1.4.2 Semantic differential tests
The aim of the second phase of the study was to investigate relations between speech intelligibility and soundscape of the native speakers of English, Polish, Arabic, and Mandarin. 15 participants per language (i.e. a total of 60) were asked to subjectively evaluate acoustic environments by answering nine questions on a five-point semantic scale, under three room acoustic conditions, in three digitally simulated multi-lingual environments. The three multi-lingual environments were an airport check-in area, a hospital reception area, and a café, i.e. three spaces where communication is crucial and
designed for each environment in order to achieve an appropriate context. Six sentences were created for each environment, and the samples were recorded by four native speakers (two males and two females) of each language in the anechoic chamber of Heriot-Watt University. The three room acoustic conditions were created digitally by adding contextually appropriate background noise and reverberation to the speech recordings. The finalised speech recordings were then presented to the participants in combination with the visuals of environments in the anechoic chamber of Heriot-Watt University, where they were asked to subjectively evaluate the audio-visual material using eleven semantic attributes (intelligibility, speech level, speech pleasantness, noisiness, annoyance, relaxation, comfort, environmental pleasantness, eventfulness, excitement, and familiarity). The results of the experiment were statistically analysed in order to identify statistically significant differences and correlations between the attributes tested.
The results obtained were used to identify the semantic attributes affecting perceived speech intelligibility of the 4 languages, to examine the differences between perceived intelligibility and actual intelligibility for various environments and room acoustic conditions, the type of environment, and the type of background noise, and to identify the semantic components affecting perceived speech intelligibility of the 4 languages.