英语翻译在流行音乐中,依据传统的风格概念进行分类,必须依靠人工赋予风格标签,这种定性水平的分类存在着很大争议,并且效率低

英语翻译
在流行音乐中,依据传统的风格概念进行分类,必须依靠人工赋予风格标签,这种定性水平的分类存在着很大争议,并且效率低下.因此,本文考虑建立数学模型,使用合适的模式识别算法,借助计算机对流行音乐进行自动分类.
本文第一个工作是流行音乐特征向量的提取.经过查阅资料得知,流行音乐之间差别主要在于高潮部分不同,平缓部分都相差不大,而高潮部分数据帧的能量大,平缓部分的数据帧能量小.因此使用数据帧能量作为特征量之一,同时可借助帧能量的大小对平缓部分的数据进行过滤；其次,数据帧能量值只是反映数值大小的特征,没有反映能量的分布特征,故而考虑将数据帧能量比作为第二个特征量.因此,流行音乐的特征向量由帧能量和帧能量比得到.为了验证特征向量提取的正确性,使用了当前一些流行歌曲进行计算验证,计算结果显示,提取的特征向量能很好的标识流行音乐,证明了特征向量提取的合理性.
本文的第二个工作是在提取特征向量的基础上,使用聚类分析的方法对流行音乐进行分类.首先使用基于最短距离的聚类算法,聚类分析的过程如下：
把每一首流行音乐单独视为一类；
根据距离最小的原则,依次选出两首音乐,并成新类；
如果其中一首音乐已归于一类,则把另一首也归入该类；如果两首音乐正好属于已归的两类,则把这两类并为一类；每一次归并,都划去该音乐所在的列与列序相同的行；
那么,经过次就可以把全部音乐归为一类,这样就可以根据归并的先后顺序作出聚类谱系图,即可完成分类.
使用该算法进行流行音乐分类的结果证明,算法具有较高的准确率,但是算法中使用的数据简单,丢失了部分信息,存在可以提升的空间.因此,本文中随后使用了模糊C均值聚类方法,对上述最短距离聚类算法进行了改进,计算实例证明,改进算法提高了对流行音乐的分类准确率,可达到84.67%.
最后,根据计算的结果,对算法的优缺点进行了总结,并对算法做了相应的推广.

In popular music, the classification of the traditional concept of style, must rely on artificial given style labels, this classification of qualitative level exist great controversy, and low efficiency. Therefore, in this paper, considering the establishment of mathematical model, using the pattern recognition algorithm is appropriate, with the help of computer automatic classification of popular music.
In this paper, the first is to extract feature vectors of pop music. Through access to information that, popular music between the difference lies mainly in the climax, flat part are similar, but the climax of data frame energy, flat part of the data frame energy. Therefore, using the data frame energy as the characteristic, also can use the frame energy size filtering of the flat part of the data; second, data frame energy value just reflect the characteristics of the numerical size, does not reflect the distribution of energy, so consider the data frame energy ratio as second features. Therefore, the feature vector of popular music from the frame energy and the frame energy ratio. In order to verify the correctness of the extracted feature vectors, using the current popular songs are validated, the results show, the extracted feature vectors can be a very good identification of pop music, proves the rationality of the extracted feature vector.
The second work is based on the feature extraction, classification of the pop music using cluster analysis method. First, using the clustering algorithm based on shortest distance, cluster analysis procedure is as follows:
Every pop song alone as a class;
According to the principle of minimum distance, in turn elect two songs, and a new class of;
If a piece of music has been attributed to a class, put another song into the class; if the two music just has to belong to class two, class two and put it into a class; each time the merge, are crossed out and sequence of the music in the same row;
Then, after you can put all the music is classified as a class, so you can according to merge to the sequence clustering dendrogram, to complete the classification.
Use of the algorithm of pop music classification results, the accuracy rate of algorithm is high, but the algorithm used in the simple data, losing some information, there can be room for improvement. Therefore, fuzzy C means clustering method is then used in this article, the shortest distance clustering algorithm is improved, the calculating example proved that, the improved classification accuracy rate of pop music, can reach 84.67%.
Finally, according to the calculation results, advantages and disadvantages of the algorithm are summarized, and the algorithm of the corresponding promotion.