Abstract
Pulmonary auscultation has been the key method to detect and evaluate respiratory dysfunctions for many years. However, auscultation with a stethoscope is a subjective process that depends on the individual’s own hearing, experience, and ability to differentiate between different sounds (Sovijarvi et al, 2000). Therefore, the computerized method for recording and analysis of pulmonary auscultative signals, being an objective way, are recently playing a more and more important role in the evaluation of patients with pulmonary diseases. Noise interference is one of the most influential factors when dealing with respiratory sound recordings. By definition of (Rossi et al, 2000), any sound not directly induced by breathing is regarded as background noise (BN). BN is divided into two types: environmental noise, which consists of continuous noise and transient noise, and nonrespiratory sounds and body sounds (muscle contraction sounds, skin friction, and heart sounds). The adaptive filtering is usually used to reduce the background noise. However, the problem of existing proposed filtering methods are either not able to minimize the interference or provides distortion which is especially undesirable for biomedical signals (Donoho, 1992).
TopMaterial And Methods
Test Dataset
Tracheal breath sounds signals from 10 healthy students of Nanyang Technological University are used as the dataset of the present study. The sample size of 10 consists of 6 females and 4 males, each producing two clips of 20-second recordings. All clips have been verified to be normal tracheal breath by Dr. Daniel Goh from National University Hospital of Singapore. At the same time, one standard preprocessed normal tracheal breath signal and two preprocessed wheeze signals from (Lehrer, 2002)(Tilkian et al, 2001)(Wilkins et al, 2004) are used together with the recorded data to test the automatic segmentation method.
Key Terms in this Chapter
Background Noise (BN): Any sound not directly induced by breathing is regarded as background noise.
Respiratory Sound: All sounds related to respiration including breath sounds, adventitious sounds, cough sounds, snoring sounds, sneezing sounds, and sounds from the respiratory muscles.
Spike Removal: A signal processing method used to remove noise of impulsive type. It involves the recognition and localization of spikes, which are short time broadband in nature.
Denoising: Denoising is any signal processing method which reconstruct a signal from a noisy one. Its goal is to remove noise and preserve useful information.
Segmentation: Segmentation means audio segmentation which is partitioning an audio stream into its homogeneous units. It refers to partition respiratory sound stream into individual inspiratory and expiratory phases.
Heartbeat Removal: A signal processing technique to remove the heart sound as acquired during sound recording, resulting a continuous pure respiratory sound stream.
Subjective Test: A subjective test session is a sequence of trials where different implementations of the proposed method are compared. Each trial provides several audio files. Each one is the result of the processing of a compared method on a reference-nondegraded file.