- مبلغ: ۸۶,۰۰۰ تومان
- مبلغ: ۹۱,۰۰۰ تومان
Sequential pattern recognition has long been an important topic of soft computing research with a wide area of applications including speech and handwriting recognition. In this paper, the performance of a novel fuzzy sequential pattern recognition tool named “Fuzzy Elastic Matching Machine” has been investigated. This tool overcomes the shortcomings of the HMM including its inflexible mathematical structure and inconsistent mathematical assumptions with imprecise input data. To do so, “Fuzzy Elastic Pattern” was introduced as the basic element of FEMM. It models the elasticity property of input data using fuzzy vectors. A sequential pattern such as a word in speech or a piece of writing is treated as a sequence of parts in which each part has an elastic nature (i.e. can skew or stretch depending on the speaker/writer’s style). To present FEMM as a sequential pattern recognition tool, three basic problems, including evaluation, assignment, and training problems, were defined and their solutions were presented for FEMMs. Finally, we implemented FEMM for speech and handwriting recognition on some large databases including TIMIT database and Dr. Kabir’s Persian handwriting database. In speech recognition, FEMM achieved 71% and 75.5% recognition rates in phone and word recognition, respectively. Also, 75.9% recognition accuracy was obtained in Persian handwriting recognition. The results indicated 18.2% higher recognition speed and 9–16% more immunity to noise in speech recognition in addition to 5% higher recognition rate in handwriting recognition compared to the HMM.
7. Conclusion and future works
The HMM has been a dominant tool in sequential pattern recognition problems such as speech and handwriting recognition. Despite the powerful modeling structure of the HMM, its low consistency with the imprecise nature of speech and handwritten data has limited its performance. Although several studies have attempted to surmount these limitations, the problem still persists as no sequential pattern recognition tool is appropriately fast, robust to noisy data, and accurate enough to be an ultimate answer. To have a fast sequential pattern recognizer with a consistent structure, when imprecise input data are given, a fuzzy sequential pattern recognition tool was presented in this paper. We focused on the basic definition ofthe elements of a pattern recognition tool by introducing the fuzzy elastic pattern. Modeling speech and handwriting with fuzzy elastic patterns gives us the ability to model the skewing or stretching of a part in the input data, which is an inherent property in speech and handwritten data. Since this tool matches the input data with the fuzzy elastic pattern and benefits from the fuzzy linguistic description, itis called the Fuzzy Elastic Matching Machine (FEMM).
In comparison with previous fuzzy logic-based approaches, the elastic matching concept was included in FEMM, leading to a more consistent model with speech and handwritten data. The major difference between FEMM and previous methods which have attempted to consider the elasticity property in the HMMby including the state duration, is a structure devoid oftoo complex relations with the aid of fuzzy set theory. Consequently, FEMM is expected to have a higher recognition speed and more immunity to noise, compared with the HMMs.