摘要
为了能够更好地识别入侵振动信号,通过研究分布式光纤振动传感器及振动信号的识别技术,根据振动信号的特点,借鉴语音信号的处理方法,对比原有基于快速傅里叶变换频谱分析算法,引入了基于Mel频率倒谱系数的识别算法。新算法从频域的角度对振动信号进行分析,提取不同环境状态下的Mel频率倒谱系数,并将其作为新的特征参量。通过实验对比分析两种算法,两者的误报率分别为27.5%和7.5%。结果表明,基于Mel频率倒谱系数的算法相比基于快速傅里叶变换的频谱分析算法,在误报率上可以降低20%甚至更多,在不漏报的前提下,显然误报率更低的基于Mel频率倒谱系数的算法更加适用于安防体系。
In order to identify the invasion of vibration signal , a recognition algorithm based on Mel frequency cepstral coefficients （ MFCC） was introduced according to the feature of vibration signal , taking processing method of voice signal as a reference and comparing with original spectrum analysis algorithm based on fast Fourier transform （ FFT） after the investigation of distributed optical fiber vibration sensor and vibration signal recognition technology .. In the new algorithm , the vibration signal was analyzed in frequency domain .MFCC under different conditions was extracted and was taken as a new characteristic parameter.After comparing the experimental results , false alarm rates of both the algorithms are 27.5%and 7.5%respectively. The results show that false alarm rate of the algorithm based on MFCC can be reduced by 20% or more compared with the algorithm based on FFT.Obviously, under the premise of not omitting, MFCC algorithm with lower false alarm rate is more suitable for security systems .
出处
《激光技术》
CAS
CSCD
北大核心
2016年第1期86-89,共4页
Laser Technology
关键词
传感器技术
分布式光纤振动传感器
振动信号
快速傅里叶变换
MEL频率倒谱系数
sensor technique
distributed optical fiber vibration sensor
vibration signal
fast Fourier transform
Mel frequency cepstral coefficient
