This paper presents a hybrid framework of feature extraction and hidden Markov modeling (HMM) for two-dimensional pattern recognition. Importantly, we explore a new discriminative training criterion to assure model compactness and discriminability. T

语音识别技术的最终目标是要让计算机能与人自由交谈。目前，连续语音识别技术正趋于成熟，语音识别也延伸出了诸多实用化的研究方向。今后，语音识别的重点将集中在自然话语识别与理解、实时语音识别和语音识别鲁棒性等方面。作为一门交叉学科，语音识别所涉及到的技术有信号处理、模式识别、概率论和信息论、发声机理、听觉机理和人工智能等。-The ultimate goal of speech recognition technology is to make computers and allowing other

机器学习和模式识别工具包spider。内容很丰富。包含svm 决策树（C45,J48）、svm、knn、adaboost、bagging、hmm（隐马尔科夫模型）、随机树（random forest）等-Machine learning and pattern recognition toolkit spider. Very rich in contents. Tree contains svm (C45, J48), svm, knn, adaboost, bagging, hmm (hidd

多种神经网络算法，包括等，可实现模式识别以及回归。-There are many algorithms in the zip,including svm, bp, crf,hmm.It can be used in the pattern recognition.