A Simple Feature Normalization Scheme for Non-native Vowel Assessment
نویسندگان
چکیده
We introduce a set of speaker dependent features derived from the positions of vowels in Mel-Frequency Cepstral Coefficient (MFCC) space relative to a reference vowel. The MFCCs for a particular speaker are transformed using simple operations into features that can be used to classify vowels from a common reference point. Classification performance of vowels using Gaussian Mixture Models (GMMs) is significantly improved, regardless of which vowel is used as the target among /A/, /i/, /u/, or /@/. We discuss how this technique can be applied to assess pronunciation with respect to vowel structure rather than agreement with absolute position in MFCC space.
منابع مشابه
Acoustic Analysis of Persian EFL Learners' Pronunciation of English Vowels
This paper reports the results of an experimental study on non-native production of English vowels. Two groups of Persian EFL learners varying in language proficiency were tested on their ability to produce the nine plain vowels of American English. Vowel production accuracy was assessed by means of acoustic measurements. Ladefoged and Maddison’s (1996) F1 F2 measurements for American English v...
متن کاملEffects of musical experience on the Thai rate-varied vowel length perception
Musical experience has been demonstrated to play a significant role in the perception of non-native speech contrasts. The present study examined whether or not musical experience facilitated the normalization of speaking rate in the perception of non-native vowel length contrasts. Musicians and non-musicians were first briefly familiarized with Thai vowel length distinctions before completing i...
متن کاملAssimilation of Final Low Back Vowel in Eghlidian Dialect
In this article, the low back vowel /A/ in word-final positions in Eghlidian dialect, one of Persian dialects, is studied. This vowel is represented phonetically as [A], [o] and [@] in different phonetic environments. Therefore many words were collected via interviewing ten native speakers so that these different alternant forms can be accounted for appropriately. Since one of the authors of th...
متن کاملSilence feature normalization for robust speech recognition in additive noise environments
In this paper, we propose a simple yet very effective feature compensation scheme for two energy-related features, the logarithmic energy (logE) and the zeroth cepstral coefficient (c0), in order to improve their noise robustness. This compensation scheme, named silence feature normalization (SFN), uses the high-pass filtered features as the indicator for speech/non-speech classification, and t...
متن کاملAutomatic evaluation of quantity contrast in non-native Norwegian speech
Computer assisted language learning (CAPT) has been shown to be effective for learning non-natives pronunciation details of a new language. No automatic pronunciation evaluation system exists for non-native Norwegian. We present initial experiments on the Norwegian quantity contrast between short and long vowels. A database of native and non-native speakers was recorded for training and test re...
متن کامل