Improvement of frequency extraction accuracy of linear predictive analysis by a combination of autocorrelation and integral procedure for noise wave shape

Abstract

AbstractThe accurate and rapid identification or detection of the electrical noise sources causing noise malfunctions in communication equipment is an important issue in determining corrective measures. In this paper, we explain an improvement in the frequency extraction accuracy of linear predictive analysis for noise waveforms in order to identify the noise sources from noise waveforms such as the audible noise generated by communication equipment like telephones and noise currents penetrating into the communication equipment. In this study, we propose an improvement of the frequency extraction accuracy of linear predictive analysis by using correlation and integration. This method preserves specific frequency components of the original noise waveforms although correlation and integration are performed. In this paper, we divide the noise waveform into finite‐length segments, correlate and integrate the time series data within a given segment, and formulate dynamic frequency analysis (time–frequency analysis) based on linear predictive analysis for the time series data in the postprocessing. We also verify the effectiveness of the proposed method by using known noise waveforms. The result of applying the proposed method to a noise waveform that convolved white noise with sine wave signals having known frequencies was clearly the ability to extract the frequencies of the sine waves included in the noise waveform even for a signal‐to‐noise ratio (SNR) of 0 dB with an error range of about 10 to 15%. This assumes that the frequencies of the sine waves included in the noise can be extracted even under the worst case SNR due to the suppression of white noise by the correlation and the emphasis of the low‐frequency components by the integration. Thus, we show that the characteristic frequencies in the noise waveform can be accurately extracted by using linear predictive analysis combining correlation and integration, and present the remaining issues. © 2004 Wiley Periodicals, Inc. Electron Comm Jpn Pt 1, 88(1): 1–11, 2005; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecja.20153