The second part of the paper consists of a list of relative formant frequencies of model vowels (the method of calculation was discussed in the first part). In this part, a test of the proposed method is designed using several new devices, such as the articulatory difference between vowels (expressed both as an integer and as a percentage), and the percentage distance between the analysed vowels and the particular model. In addition, a few simple statistics such as median, mode and arithmetic mean are used. The test itself consists of recognizing new vowels as pronounced by three different phoneticians. The calculations lead to the conclusion that recognition of vowels carried out with the use of the proposed method produces very accurate results which, most importantly, do not deviate in most instances from the expected values (the difference between the recognized vowel and the vowel described by the speaker did not exceed 40%). In conclusion, this part of the paper describes the advantages and limitations of the proposed method, and suggests possible solutions which may help improve the results in the future.
2
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
The first part of the paper proposes a method of studying (or, more precisely, of recognizing) oral vowels, using the so-called relative formant frequencies. First, we describe the traditional method of acoustic study of oral vowels, together with its limitations. Next, we introduce the assumptions of the proposed method where, unlike in the traditional approach, a set of mixed models is used. The comparison, instead of being reduced to establishing whether two sets of vowels match, requires a distance to be calculated between each vowel and each model. The paper introduces the notion of relative formant frequency (denoted f') which was used in the comparisons to minimize the impact of individual articulatory variation on the outcome of the analysis. It was de!ned as the proportion between, on the one hand, the difference between the absolute frequency of the given formant, and its minimal value observed in actual pronunciations, and on the other hand, the difference between the minimal and the maximal frequency of the formant, as observed in pronunciations. In order to establish the f' of model vowels, we used the cardinal vowels pronounced by several British phoneticians (including Daniel Jones).
JavaScript jest wyłączony w Twojej przeglądarce internetowej. Włącz go, a następnie odśwież stronę, aby móc w pełni z niej korzystać.