Speech Science - Science topic

TO: Ghadah Marie

Thank you very much.

Hi Andrey. Although my wife is Chinese, I cannot contribute to this subject. I do not speak Yoruba, either. However, considering your incredibly wide range of interests in many sciences, I suggest you to take a look at my website named CorrectingWorldHistory. Also, I realized that many Philippine languages and some western African languages (west of the Yoruba speakers) have almost the same word for "good" (mabute, mbuti, etc.). Perhaps there are more similarities. If one could prove such relationship between the black (Melanesian) nations of the Philippines (or the Aborigines of Australia) and some of the African languages we may suggest that the same word existed in both continents before, say, 40,000 years B.P. Thus, the basic words of these languages are perhaps not only a few thousand years old but perhaps 50 thousand years old. I congratulate to your research papers and results. I am originally Hungarian who lives in Canada since 1976. Have a great weekend!!!

It is one of the old issues in a speech recognition research field. That is on the relationship between any speech enhancement technique and the classification accuracy.

As far as I know, both the MSE and PSNR are frequently used for improving the quality of the input. They are known as useful in reducing WER. However, the relationship with recognition accuracy is not directly proportional. 

Enhancing a noisy signal in terms of MSE or PSNR means that you may have a good quality of the input but there is a risk. Sometimes, unexpected artifacts are produced by the speech enhancement techniques and WER can be increased in the worst case.

So, in phonemic classification task, matched condition is more crucial. And in the case of mismatched condition between train and test, MSE and PSNR are somewhat related to WER, but not directly. It is a case-by-case study.    

Hello all

I really appreciate all your support. 

Regards

Fateme

for a huge free dataset look at:

McKenzie (2010) explored more than 500 Japanese university students’ attitudes towards six ‘varieties’, and relevant articles are McKenzie (2008a, 2008b).  More recent attitude-related studies of his are published and available on his RG account.

McKenzie, Robert M. 2008a. The role of variety recognition in Japanese university students’ attitudes towards English speech varieties. Journal of Multilingual and Multicultural Development 29(2). 139–153.

McKenzie, Robert M. 2008b. Social factors and non-native attitudes towards varieties of spoken English: A Japanese case study. International Journal of Applied Linguistics 18(1). 63–88.

McKenzie, Robert M. 2010. The social psychology of English as a global language: Attitudes, awareness and identity in the Japanese context. Dordrecht: Springer.

Jenkins (2007) explored more than 300 English teachers' attitudes towards diverse accents, and a relevant article is Jenkins (2009).  Also, Jenkins (2014: Ch.7) revealed international students’ orientations towards diversity in English.

Jenkins, Jennifer. 2007. English as a Lingua Franca: Attitude and identity. Oxford: Oxford University Press.

Jenkins, Jennifer. 2009. English as a lingua franca: Interpretations and attitudes. World Englishes 28(2). 200–207.

Jenkins, Jennifer. 2014. English as a Lingua Franca in the international university: The politics of academic English language policy. London: Routledge.

Hope this helps.

Hi Bettina,

if you have enough data from your subjects, z-scoring, but separately for each phoneme class, may be an option. I've used phoneme-specific z-scores for recognizing prosodic boundaries and pitch accented syllables. This takes into account Hartmut's finding, mentioned by Susanne above, that different phonemes are affected differently. Nick Campbell, by the way, came up with the term elasticity -- phonemes differ in their elasticity, and he introduced the phoneme-specific z-scores to model phoneme durations in synthesis in a paper in 1992.

Antje

Hi Barbara,

Have you read their follow-up paper? I believe it gives you the information you are looking for.

Ariel

There are two types of features of a speech signal:

I think, you have to detect both, voicing level of speech signal and unvoiced portions and the time depency of the transition.