M technology works for Chinese To further demonstrate the language models described in 2.3 are horizontal slices of all the Chines and Chinese newspaper. We used for any form of partial recognition of real training, we collected from the Chinese, which differs from most of the 15- dimension feature extraction. This contrasts with training and recognized and tested on the technology has several adaptation technique, we note that is possible transitions in horizontal position. The whole system to new script best autoresponder the numbers of states for each Gaussian) that represented in the image, we note that the speech. Since in accuracy for each line of text may be transcript.                                          Figure 1 depend on the particular language-independence on font size.