Automatic Speech Recognition (ASR) is a well developed technical area, conjoining technologies such as DSP, Machine Learning, HS Codesign, and Linguistics. For scientific research ASR is an invaluable source of ideas and solutions integrated in one important product. Unfortunately, being a business interest, advanced ASR technologies are hidden by companies. The open literature is focused round the Hidden Markov Models (HMM) which was developed earlier in 1960’s and then incorporated into the HMM as an attractive technique. Open ASR systems with HMM report surprisingly low level recognition – 75%. After the beginning of HMM era many things changed in HS area – now very large memories are available, even the portable devices use multiprocessors with high computational power. It is to analyze what is achieved and what reminds to be understood with ASR in new situation and probably this helps in increasing the accuracy of recognition. Our target after this analysis will be the setting up of advanced ASR system (for Armenian language). In fact the system will be learnable, when large linguistic corpora are available, so that it can be multilingual. The feature ASR technology that is visible today is related to the Sequence Data Mining (SDM) technique. Being part of the structural data mining this technique is already studied and developed. Studies involve the topics of LCS, gene alignment and similar, but speech signal analysis provide a specific situation that is to be studied deeply, with consequent optimization and implementation. All the related thoughts are openly presented in the text below. Future work will enlarge the given ongoing prototype with means of advanced SDM and with interfaces.