基于频域匹配滤波的音频信号分析与识别

Analysis and Recognition of Audio Signal Based onFrequency Domain Matching Filtering

  • 摘要: 音频信号识别在音乐数据库检索技术和计算机自动谱曲等领域具有重要的实用价值。该文以钢琴音阶信号为例,侧重对采集的音频信号进行频域的分析和匹配,实现音频信号的识别。首先,利用Goertzel算法得到每个标准音的频谱序列,并将得到的序列能量归一化为标准模板;其次,将待识别的音频信号的频谱序列与每个标准音的模板一一匹配;最后,由最佳匹配结果找到对应的乐音音符。通过对钢琴标准音频文件、手机钢琴演奏软件以及真实的钢琴演奏录音的检测表明,该方法具有较高的识别准确度和良好的抗噪性能。

     

    Abstract: Audio signal recognition is of great practical value in music database retrieval technology and computer automatic music composing.Taking the piano scale signals as an example,this paper focuses on the analysis,matching and identification of the collected audio signals in frequency domain,so as to realize the recognition of the audio signals.Firstly,the spectra of standard music notes are obtained by using the Goertzel algorithm,and the obtained sequence energy is normalized to standard template.Secondly,the spectrum of the audio signal to be identified is calculated and matched with the template of each standard sound.Finally,the sound notes are found by the best matching results.Through the test of the piano standard audio file,the mobile piano playing software and the real piano performance recording,the method has high recognition accuracy and good anti-noise performance.

     

/

返回文章
返回