Towards capturing fine phonetic location in speech using recurrence dynamic feature

The ultimate goal of our research is to develop a new speech dynamic feature that is able to capture the effects of fine-grained acoustic location on speech recognition behavior. As part of this work we are investigating recurrence quality analysis (RQA) and cross recurrence plot (CRP) approach. In the experiments reported here, we analyzed the Chinese vowel and vocabulary using CRP. Experimental results demonstrate that this method is effective in location; it could be beneficial to develop speech recognition.