Showing papers by "Yaniv Zigel published in 2013"

Automatic detection of whole night snoring events using non-contact microphone.

Abstract: Objective Although awareness of sleep disorders is increasing, limited information is available on whole night detection of snoring. Our study aimed to develop and validate a robust, high performance, and sensitive whole-night snore detector based on non-contact technology.

OSA severity assessment based on sleep breathing analysis using ambient microphone

Abstract: In this paper, an audio-based system for severity estimation of obstructive sleep apnea (OSA) is proposed. The system estimates the apnea-hypopnea index (AHI), which is the average number of apneic events per hour of sleep. This system is based on a Gaussian mixture regression algorithm that was trained and validated on full-night audio recordings. Feature selection process using a genetic algorithm was applied to select the best features extracted from time and spectra domains. A total of 155 subjects, referred to in-laboratory polysomnography (PSG) study, were recruited. Using the PSG's AHI score as a gold-standard, the performances of the proposed system were evaluated using a Pearson correlation, AHI error, and diagnostic agreement methods. Correlation of R=0.89, AHI error of 7.35 events/hr, and diagnostic agreement of 77.3% were achieved, showing encouraging performances and a reliable non-contact alternative method for OSA severity estimation.

Apparatus and method for diagnosing sleep quality

Abstract: A method of distinguishing sleep period states that a person experiences during a sleep period, the method comprising: using a non-contact microphone to acquire a sleep sound signal representing sounds made by a person during sleep; segmenting the sleep sound signals into epochs; generating a sleep sound feature vector for each epoch; providing a first model that gives a probability that a given sleep period state experienced by the person in a given epoch exhibits a given sleep sound feature vector; providing a second model that gives a probability that a first sleep period state associated with a first epoch transitions to a second sleep period state associated with a subsequent second epoch; and processing the feature vectors using the first and second models to determine a sleep period state of the person from a plurality of possible sleep period states for each of the epochs.

Separating clinically relevant sources of electrical activity in ecg signals

Abstract: This invention provides a fully automatic method and a system for detecting and classifying cardiac arrhythmias from a surface ECG record. The method defines four relevant parameters whose values are extracted from said ECG record. Clinically relevant conclusions are assigned to various combinations of the obtained values of said parameters. The method can be employed for detecting fetal QRS complexes from abdomen ECG of a pregnant woman.

Noninvasive fetal QRS detection using a linear combination of abdomen ECG signals

Abstract: The fetal ECG can serve as a tool for fetal distress detection. However, the abdominal ECG of a pregnant woman contains mainly the maternal ECG and a relatively small amplitude fetal ECG signal, contaminated by various noises. As part of the 2013 PhysioNet/CinC Challenge, this study aimed to develop an algorithm for noninvasive fetal QRS detection. The proposed algorithm is mainly based on fetal ECG source signal enhancement using a modified linear combiner. After initial noise reduction, the maternal QRS complexes are detected. Then, fetal QRS candidates are found. For each candidate, a Gaussian-like synthetic fetal QRS signal is created. This signal is considered an observation signal for a modified linear combiner. The 4 filtered abdomen ECG signals then undergo maternal ECG cancellation and serve as reference signals in this linear combiner; hence by finding the appropriate weight coefficients, their linear combination is forced to converge to a signal that represents the fetal QRS complexes solely. The method was developed using the entire 75 1minute-long abdomen ECG training set, and its evaluation on the 100 abdomen ECG test set led to scores of 262.076 for event 4 (fetal heart rate measurement) and 27.848 for event 5 (fetal RR interval measurement).

Effective Model Representation by Information Bottleneck Principle

Abstract: The common approaches to feature extraction in speech processing are generative and parametric although they are highly sensitive to violations of their model assumptions Here, we advocate the non-parametric Information Bottleneck (IB) IB is an information theoretic approach that extends minimal sufficient statistics However, unlike minimal sufficient statistics which does not allow any relevant data loss, IB method enables a principled tradeoff between compactness and the amount of target-related information IB's ability to improve a broad range of recognition tasks is illustrated for model dimension reduction tasks for speaker recognition and model clustering for age-group verification