552 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 57, NO. 3, MARCH 2010 Analysis and Modeling of Snore Source Flow With Its Preliminary Application to Synthetic Snore Generation Andrew Keong Ng ∗ , Student Member, IEEE, and Tong San Koh Abstract—With the emerging use of snore properties for clinical purposes, there is a need to understand the characteristics of snore source flow (SF)—the acoustic source in snore production. This pa- per attempts to analyze and model both SF and its derivative (SFD), along with its preliminary application to the generation of synthetic snores. SFs and SFDs were extracted from natural snores via an iterative adaptive inverse filtering approach, and subsequently pa- rameterized into various time- and amplitude-based parameters to quantify the oscillatory maneuvers of snore excitation source (ES). The SF and SFD waveforms were also, respectively, modeled using the first and second derivatives of the Gaussian probabil- ity density function. Subjective and objective measures, including paired comparison score and sum-of-squared error, were assessed to appraise the performance of SFD model in producing natural- sounding snores. Results consistently show that: 1) the shapes of SF pulse are different among snores and can be associated with the dynamic biomechanical properties (e.g., compliance and elas- ticity) of ES; 2) changes to the SF or SFD pulse shape can affect the snore properties, both acoustically and perceptually; and 3) the proposed SFD model can generate close-to-natural sounding snores. Further research in this area can potentially yield valuable benefits to snore-oriented applications. Index Terms—Gaussian function, Mexican hat wavelet, obstruc- tive sleep apnea (OSA), signal modeling, signal parameterization, snore excitation source (ES), snore source flow (SF), snore SF derivative (SFD), snore synthesis, snoring. I. INTRODUCTION S NORES are respiratory sounds generated by vibrations of soft tissues (e.g., soft palate, uvula, tonsils, tongue base, epiglottis, and lateral pharyngeal walls) [1]–[3] and/or turbu- lence of airflow at constrictions in the upper airway (UA). Acoustical properties of snores have increasingly demonstrated their clinical usefulness in locating snoring sites for surgical planning [4], [5], predicting surgical treatment outcomes for snoring [6], [7], and identifying snorers with or without ob- structive sleep apnea (OSA, a prevalent sleep-related breathing disorder that is usually accompanied by snoring) [8]–[16]. Considerable attempts have been made to unfold the clinical value of snores [4]–[16], with a prevailing recognition that the nature and extent of variations in snore source flow (SF), which Manuscript received May 5, 2009; revised July 10, 2009, August 24, 2009, and September 26, 2009. First published October 20, 2009; current version pub- lished February 17, 2010. Asterisk indicates corresponding author. ∗ A. K. Ng is with the School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore 639798 (e-mail: andrewkng@ pmail.ntu.edu.sg). T. S. Koh is with the School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore 639798 (e-mail: etskoh@ntu.edu.sg). Digital Object Identifier 10.1109/TBME.2009.2034139 is the volume velocity of airflow at snore excitation source (ES), plays an influential role in the diagnostic outcome [1]–[3]. A feasible rationale behind its importance in decision making is that a snore can be approximately modeled as an output of a linear time-invariant filtering operation between the ES, UA, and lip radiation [2], [9], [15]–[17], analogous to the production of speech [18]–[20]. Explicitly, aerodynamic forces at constric- tions in the UA excite ES vibrations, modulating the steady airflow to a pulsating airflow (snore SF). The SF, denoted by u(t), is then modified by the UA acting as an acoustic filter with impulse response of v(t). The volume velocity output of the UA is eventually radiated from the lips, which can be regarded as a differentiator [18]–[20], giving rise to a snore sound pressure signal x(t) ≈ d[u(t) ∗ v(t)]/dt = [du(t)/dt] ∗ v(t), where du(t)/dt is the snore SF derivative (SFD) that combines the effects of SF and lip radiation and constitutes the input to the UA. Direct measurement of snore SF is a formidable challenge as it engages in invasive monitoring of physiological process of snore generation during sleep. Hence, the knowledge about SF remains incomplete, yet it is undeniably essential for nu- merous snore-oriented applications (e.g., snore signal analysis, synthesis, coding, and recognition) to augment clinical decision making. Most of what is known comes from theoretical and structural models that illustrate snoring mechanisms [21]–[27]. These models no doubt provide an in-depth understanding of the interactions between airflow and soft tissues, together with interplay factors leading to snoring and OSA; they need full details of the UA structure and function, which one does not usually possess because of its geometric complexity. Further- more, no emphasis is placed on the perceptual significance of fluid–structure interactions. This paper therefore aims to offer greater insights into snore SF by analyzing and modeling natural (without anesthesia) snore SFs and SFDs, with an underlying hypothesis that the tempo- ral and spectral characteristics of both SF and SFD contain rich information on ES dynamics and affect the attributes of snores, both acoustically and perceptually. To achieve the ob- jectives, we extracted SFs and SFDs directly from nocturnal snores, parameterized, and modeled a common SF waveform and SFD waveform, and subsequently synthesized snores using the proposed SFD model. Subjective and objective measures, in terms of paired comparison score and sum-of-squared er- ror, were evaluated to appraise the performance of the SFD model in generating natural-sounding snores. To the best of our knowledge, this is the first attempt to analyze and model snore SF or SFD, accompanied by its preliminary application to 0018-9294/$26.00 © 2009 IEEE