介绍了孤立词语音识别系统，针对片上系统进行了语音识别算法的选择。
This paper introduces the isolated word speech recognition for system on chip (SoC).
根据语音识别算法原理，对语音信号的预处理、特征提取、模式匹配做了分析。
According to the voice identification arithmetic theory, the pretreatment of voice signal, the character distilling and pattern matching is analyzed.
快速模板匹配(FTM)是一种非常实用高效的实时语音识别算法，尤其适用于命令识别系统。
Fast Template Matching (FTM) is a very practical, effective realtime speech recognition algorithm, especially suitable for command recognition system.
本文在简单介绍词组语音识别之后给出了一个基于连续语音识别算法和词树约束的汉语词组语音识别方法。
After an overview of phrase recognition, this paper proposes a Chinese phrase recognition method based on continuous speech recognition algorithm and phrase-tree constraint.
结果表明：所设计的语音识别算法有很高的识别率，能减小或者消除噪声所带来的训练模型和测试语音之间的失配。
The results show that the speech recognition algorithm has high recognition rate, can reduce or eliminate noise caused by the training model and the mismatch between the speech test.
这些经典的语音识别算法是基于模式匹配的方法实现，这使其在语音分类中的性能受到挑战，尤其是在对相近的中文姓名语音的分类中。
These classic algorithms are based on pattern matching, this causes decline of the performance especially in classification of similar Chinese name speech patterns.
然后，他们将庞大的声响数据库发送给计算机，以验证机器学习是否能帮助他们理解这些问题，比如使用用于人类语音识别的算法。
They then fed their huge database of calls to computers, to test whether machine learning could help make sense of them, using algorithms like the ones used for human speech recognition.
苹果Siri服务打开了我们的想象力；但谷歌拥有令人赞叹的语音识别技术、算法和卓越的平台实力，来完成这些事情。
Siri has opened our imagination; but Google has amazing voice recognition, algorithmic and platform strength to accomplish these things.
然后深入研究了语音识别的基本算法与技术。
Then the article expatiated the algorithm and technology of speech recognize.
系统硬件的核心芯片采用普遍使用的16位定点DSP芯片TMS320 VC 5416，软件上主要采用动态时间规整算法来实现语音识别。
TMS320VC5416 16-bit fixed-point DSP is adopted in its core processor of the hardware. Dynamic Time Warping (DTW) algorithm is chiefly used to implement the software.
本文提出了一种孤立词语音识别系统中基于后验概率差值的拒识算法。
This paper proposes an algorithm based on likelihood difference for isolated-word speech recognition system.
提出了一种基于SDCN算法的鲁棒性语音命令识别。
A robust voice command recognition based on SDCN algorithm was presented.
实验表明，MAP算法可以有效地降低汉语数码语音识别对被适应人的误识率，而且对非自适应人性能影响很小。
The experiment shows that the MAP algorithm can considerably reduce the error rate of Mandarin digital speech recognition on the adapted speaker and cause little influence on the non-adapted speakers.
第二部分讲述语音识别系统中的有关演算法，包括模式分类，查找演算法，随机模型，语言模型等技术。
Part II describes algorithmic aspects of speech recognition systems including pattern classification, search algorithms, stochastic modelling, and language modelling techniques.
完整的连续语音识别系统主要包括四个部分：特征提取，声学模型，语言模型和搜索算法，本文就是根据这四个部分展开的。
A whole continuous speech recognition system includes four parts: feature extraction, acoustic model, language models and search algorithms, and the thesis is carried out according to them.
词边界检测误差是语音识别中产生错误的主要原因之一，常规的检测算法在低信噪比尤其在背景噪声能量可变的环境下不能有效工作。
A major cause of errors in speech recognition is the inaccurate detection of word bound-ary. Conventional methods can not work well in the condition of low SNR and variable background noise level.
然后用MATLAB实现了这两种算法，并进行了大量的孤立词语音识别实验。
The two algorithms are implemented using MATLAB, and experiments of isolated-word recognition are done.
实验表明，LOPDI算法能够显著提高语音识别系统对加性噪声的鲁棒性。
Experimental result shows that LOPDI algorithm can greatly increase automatic speech recognition system's robustness against additive noise.
实验表明，MAP算法可以有效地降低汉语数码语音识别对被适应人的误识率，而且对非自适应人性能影响很小。
The experiment shows that the MAP algorithm can considerably reduce the error rate of Mandarin digital speech recognition on the adapted speaker and cause litt...
同时，根据语音信号不连续的特点，压缩算法具有静音识别功能，进一步提高了语音信号的压缩率。
According to the characteristic of discontinuous voice signal, this compression algorithm provides the function of voice active detection and improves the compress ratio of voice signal.
针对高噪音环境中的语音识别问题，提出一种利用模糊神经网络进行语音数据融合的新算法。
A new algorithm of employing the fuzzy neural network is proposed to realize speech data fusion for speech recognition under high noisy condition.
为了满足超大词表语法的识别任务在嵌入式语音识别系统上的应用，提出了一种高效的双层图搜索算法。
For the application of large vocabulary word lists to embedded speech recognition, an efficient two-layer graph search algorithm is proposed.
随着待识别语音信号数量的增多，该算法优势更加明显。
The advantage of the improved algorithm will be more obvious with the increasing number of the voice signal to be recognized.
将基于连续隐含Markov模型语音 识别算法中占系统总运算量的50%以上的Mahalanobis距离 计算，映射为硬件实现的 模块。
The mapped algorithm, called the Mahalanobis distance, handles about 50% of its computational load in the overall speech recognition algorithm using a continuous hidden Markov model(CHMM).
将基于连续隐含Markov模型语音 识别算法中占系统总运算量的50%以上的Mahalanobis距离 计算，映射为硬件实现的 模块。
The mapped algorithm, called the Mahalanobis distance, handles about 50% of its computational load in the overall speech recognition algorithm using a continuous hidden Markov model(CHMM).
应用推荐