基于端点检测的时频峰值滤波动液面提取技术

油套环空中会产生各种噪声，使测得的液面反射信号非常复杂，真实的液面反射波位置因受到干扰而无法准确辨识。采用时频峰值滤波（TFPF）技术结合语音信号处理中的端点检测（VAD）方法可对动液面波进行有效提取。VAD-TFPF技术先采用短时能量和过零率的双门限VAD方法对声波法测油井信号进行划分，判断出有效信号数据段和接近于零值的数据段，然后采用不同窗长的TFPF分别对两种数据段进行滤波处理。通过对不同噪声强度下的实测数据进行滤波实验与分析，可知该项新技术较之于小波阈值滤波方法对动液面波的辨识能力更强，无论是对背景噪声的压制还是对有效波的提取都表现出更优越的性能。

Working Fluid Level Extraction Technique Using Time-frequency Peak Filtering Based on Voice Activity Detection

Several types of noises are generated in an oil jacket annulus, which complicates the fluid level reflection signal measurement. The real position of the fluid level reflection wave could not be accurately identified owing to interference. Thus, the time-frequency peak filtering (TFPF) method is applied, along with the voice activity detection (VAD) method in voice signal processing, to perform effective extraction of the working fluid level wave. The VAD-TFPF method first uses the short-time energy and double-threshold VAD method with a zero-crossing rate, divides the oil well signal measured using the acoustic method, and determines the effective signal data segment and data segment close to zero; then, TFPF with different window lengths is applied to perform filtering to two types of data segments. The filtering experiment and analysis are carried out to measured data under different noise intensities; the results indicate that this method has a stronger identification ability toward working fluid level waves than the wavelet threshold filtering method and is preferable for background noise suppression or for effective wave extraction.