Bài giảng Biomedical signal processing and modeling - Analysis of Nonstationary and Multicomponent Signals
Characterization of Nonstationary Signals and Dynamic Systems
Mean:
• The short-time mean represents the average or DC level of the signal in the analysis window.
• Variation of the mean from one window to another is usually an indication of the presence of a wandering baseline or low-frequency artifact.
• However, the mean is not an important measure in most signals, and it is typically blocked at the data-acquisition stage via capacitive coupling and/or a highpass filter.
• Furthermore, since a DC level carries no sound or vibration information, its removal is of no consequence in the analysis of signals such as heart sounds, speech, VAG, and the VMG.
Variance:
• The variance is high in regions of high signal variability (swings or excursions) about the mean.
• The variance is low in the regions related to the fricatives or unvoiced-speech segments in the signal where the amplitude swing is small, in spite of their high-frequency nature.
• Since the mean of the signal is zero, the variance is equal to the mean square value and represents the average power level in the corresponding signal windows.
• This signal is nonstationary in its variance.
Measures of activity:
• Turning points, ZCR, turns count.
• They characterize signal variability and complexity in different ways.
• This signal is nonstationary in terms of the number of turning points.
Mean:
• The short-time mean represents the average or DC level of the signal in the analysis window.
• Variation of the mean from one window to another is usually an indication of the presence of a wandering baseline or low-frequency artifact.
• However, the mean is not an important measure in most signals, and it is typically blocked at the data-acquisition stage via capacitive coupling and/or a highpass filter.
• Furthermore, since a DC level carries no sound or vibration information, its removal is of no consequence in the analysis of signals such as heart sounds, speech, VAG, and the VMG.
Variance:
• The variance is high in regions of high signal variability (swings or excursions) about the mean.
• The variance is low in the regions related to the fricatives or unvoiced-speech segments in the signal where the amplitude swing is small, in spite of their high-frequency nature.
• Since the mean of the signal is zero, the variance is equal to the mean square value and represents the average power level in the corresponding signal windows.
• This signal is nonstationary in its variance.
Measures of activity:
• Turning points, ZCR, turns count.
• They characterize signal variability and complexity in different ways.
• This signal is nonstationary in terms of the number of turning points.
Bạn đang xem 20 trang mẫu của tài liệu "Bài giảng Biomedical signal processing and modeling - Analysis of Nonstationary and Multicomponent Signals", để tải tài liệu gốc về máy hãy click vào nút Download ở trên.
File đính kèm:
bai_giang_biomedical_signal_processing_and_modeling_analysis.pdf
Nội dung text: Bài giảng Biomedical signal processing and modeling - Analysis of Nonstationary and Multicomponent Signals
- Nguyễn Công Phương BIOMEDICAL SIGNAL PROCESSING AND MODELING Analysis of Nonstationary and Multicomponent Signals
- Contents I. Introduction II. Concurrent, Coupled, and Correlated Processes III. Filtering for Removal of Artifacts IV. Detection of Events V. Analysis of Waveshape and Waveform Complexity VI. Frequency Domain Characterization VII.Modeling Biomedical Systems VIII.Analysis of Nonstationary and Multicomponent Signals IX. Pattern Classification and Diagnostic Decision sites.google.com/site/ncpdhbkhn 2
- Analysis of Nonstationary and Multicomponent Analysis 1. Introduction 2. Time-variant Systems 3. Characterization of Nonstationary Signals and Dynamic Systems 4. Fixed Segmentation 5. Adaptive Segmentation 6. Use of Adaptive Filters for Segmentation 7. Wavelet and Time-frequency Analysis 8. Separation of Mixtures of Signals sites.google.com/site/ncpdhbkhn 3
- Introduction • A stationary (or homogeneous) signal is one that possesses the same statistical measures for all time, or at least over the duration of observation. • Most biomedical signals, being manifestations of dynamic systems and pathophysiological processes, are nonstationary (or heterogeneous). • A multicomponent signal is the one that has delineated concentrations of power or energy in the time-frequency (TF) plane, with each such portion possibly related to a component of the signal. • Problems: – Develop methods to study the dynamic characteristics of nonstationary and multicomponent biomedical signals. – Propose schemes to apply the well-established Fourier transform and AR-modeling techniques to analyze and parameterize nonstationary and multicomponent signals. sites.google.com/site/ncpdhbkhn 4
- Analysis of Nonstationary and Multicomponent Analysis 1. Introduction 2. Time-variant Systems 3. Characterization of Nonstationary Signals and Dynamic Systems 4. Fixed Segmentation 5. Adaptive Segmentation 6. Use of Adaptive Filters for Segmentation 7. Wavelet and Time-frequency Analysis 8. Separation of Mixtures of Signals sites.google.com/site/ncpdhbkhn 5
- Time-variant Systems P Q =− −+ − yn[] aynkk [] G bxnl l [] k=1 l = 0 • Time-invariant system: – ak and bl are constant. – (therefore) its pole(s) & zero(s) stay fixed for all time. • Nonstationary/dynamic system: – ak and bl vary with time. – E.g., the adaptive LMS & RLS filters. – Its frequency response & impulse response vary with time. sites.google.com/site/ncpdhbkhn 6
- Analysis of Nonstationary and Multicomponent Analysis 1. Introduction 2. Time-variant Systems 3. Characterization of Nonstationary Signals and Dynamic Systems 4. Fixed Segmentation 5. Adaptive Segmentation 6. Use of Adaptive Filters for Segmentation 7. Wavelet and Time-frequency Analysis 8. Separation of Mixtures of Signals sites.google.com/site/ncpdhbkhn 7
- Characterization of Nonstationary Signals and Dynamic Systems (1) 1.5 Mean : 1 • The short-time mean represents the average or DC level of the signal in 0.5 the analysis window. 0 • Variation of the mean from one -0.5 window to another is usually an indication of the presence of a -1 -1.5 wandering baseline or low-frequency 0 100 200 300 400 500 artifact. • However, the mean is not an important measure in most signals, and it is 1 typically blocked at the data- N = 8 acquisition stage via capacitive 0.5 coupling and/or a highpass filter. • Furthermore, since a DC level carries 0 no sound or vibration information, its removal is of no consequence in the analysis of signals such as heart -0.5 sounds, speech, VAG, and the VMG. -1 0 100 200 300 400 500 sites.google.com/site/ncpdhbkhn Time 8
- Characterization of Nonstationary Signals and Dynamic Systems (2) 0.3 Variance : 0.2 • The variance is high in regions of high signal variability (swings or 0.1 excursions) about the mean. 0 • The variance is low in the regions -0.1 related to the fricatives or -0.2 unvoiced-speech segments in the -0.3 1 2 3 4 5 6 signal where the amplitude swing 10 4 is small, in spite of their high- frequency nature. 10 -3 N = 8 • Since the mean of the signal is 12 zero, the variance is equal to the 10 mean square value and represents the average power level in the 8 corresponding signal windows. 6 • This signal is nonstationary in its 4 variance. 2 1 2 3 4 5 6 sites.google.com/site/ncpdhbkhn Time 9 10 4
- Characterization of Nonstationary Signals and Dynamic Systems (3) Measures of activity : 0.2 0 • Turning points, ZCR, -0.2 turns count. 1 2 3 4 5 6 • 10 4 They characterize signal 60 N = 100 variability and 40 complexity in different 20 ways. 0 1 2 3 4 5 6 • 10 4 This signal is 10 -3 12 N = 8 nonstationary in terms of 10 8 the number of turning 6 4 points. 2 1 2 3 4 5 6 sites.google.com/site/ncpdhbkhn Time 10 10 4