在孤立词语音识别中，最为简单有效的方法是采用DTW(Dynamic Time Warping，动态时间归整)算法，该算法基于动态规划(DP)的思想，解决了发音长短不一的模板匹配问题，是语音识别中出现较早、较为经典的一种算法，用于孤立词识别。HMM算法在训练阶段需要提供大量的语音数据，通过反复计算才能得到模型参数，而DTW算法的训练中几乎不需要额外的计算。所以在孤立词语音识别中，DTW算法仍然得到广泛的应用。
文 件 列 表
com util timeseries matrix lang dtw AbstractionTest.java BandTest.java DtwTest.java FastDtwTest.java Test.java
SHOW FULL COLUMNS FROM `jrk_downrecords` [ RunTime:0.001301s ]
SELECT `a`.`aid`,`a`.`title`,`a`.`create_time`,`m`.`username` FROM `jrk_downrecords` `a` INNER JOIN `jrk_member` `m` ON `a`.`uid`=`m`.`id` WHERE `a`.`status` = 1 GROUP BY `a`.`aid` ORDER BY `a`.`create_time` DESC LIMIT 10 [ RunTime:0.093912s ]
SHOW FULL COLUMNS FROM `jrk_tagrecords` [ RunTime:0.001227s ]
SELECT * FROM `jrk_tagrecords` WHERE `status` = 1 ORDER BY `num` DESC LIMIT 20 [ RunTime:0.001921s ]
SHOW FULL COLUMNS FROM `jrk_member` [ RunTime:0.001289s ]
SELECT `id`,`username`,`userhead`,`usertime` FROM `jrk_member` WHERE `status` = 1 ORDER BY `usertime` DESC LIMIT 10 [ RunTime:0.004017s ]
SHOW FULL COLUMNS FROM `jrk_searchrecords` [ RunTime:0.001163s ]
SELECT * FROM `jrk_searchrecords` WHERE `status` = 1 ORDER BY `num` DESC LIMIT 5 [ RunTime:0.003396s ]
SELECT aid,title,count(aid) as c FROM `jrk_downrecords` GROUP BY `aid` ORDER BY `c` DESC LIMIT 10 [ RunTime:0.016993s ]
SHOW FULL COLUMNS FROM `jrk_articles` [ RunTime:0.001565s ]
UPDATE `jrk_articles` SET `hits` = 1 WHERE `id` = 42463 [ RunTime:0.001609s ]
