A novel multipitch measurement algorithm for acoustic signals of moving targets. (15th December 2016)

• Record Type: Journal Article
• Authors: Huang, Jingchang; Guo, Feng; Zu, Xingshui; Li, Haiyan; Liu, Huawei; Li, Baoqing
• Is Part Of: Mechanical systems and signal processing. Volume 81(2016)
• Page Start: 419
• Page End: 432
• Publication Date: 2016-12-15
• Subjects: Acoustic signal analysis -- Fundamental frequency measurement -- Harmonics -- Moving targets -- Multi-impulse-train -- Symmetric average magnitude sum function; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2016.02.048
• Languages: English
• ISSNs: 0888-3270

Abstract

Abstract: In this paper, a novel multipitch measurement (MPM) method is proposed for acoustic signals. Starting from the analysis of moving targets׳ acoustic signatures, a pitch-based harmonics representation model of acoustic signal is put forward. According to the proposed harmonics model, a modified greatest common divisor (MGCD) method is developed to obtain an initial multipitch set (IMS). Subsequently, the harmonic number vector (HNV) associated with the IMS is determined by maximizing the objective function formulated as a multi-impulse-train weighted symmetric average magnitude sum function (SAMSF) of the observed signal. The frequencies of SAMSF are determined by the target acoustic signal, the periods of the multi-impulse-train are governed by the estimated IMS harmonics and the maximization of the objective function is figured out through a time-domain matching of periodicities of the multi-impulse-train with that of the SAMSF. Finally, by using the obtained IMS and its HNV, a precise fundamental frequency set is achieved. Evaluation of the algorithm performances in comparison with state-of-the-art methods indicates that MPM is practical for the multipitch extraction of moving targets. Highlights: A pitch-based signal representation model is put forward. The proposed multipitch measurement algorithm is compact and accurate. Experiments achieve excellent performances. Easy to apply the proposed scheme to other harmonic signals.