Tutorial & Demo
The GreatMiner is designed to serve researchers at signal processing. It lays all data in a multi-frame layout for joint analysis, while allows to investigate at individual signal in deep. It is also used to pre-treat data for being used by third-party software.
- Multiple Signals Automatic LayoutWhen you load one or multiple signals for processing by GreatMiner, the signals are saved into an in-memory matrix until another loading operation is processed and new signals take place of the old ones. The signals at the in-memory matrix are always displayed with a multi-frame layout. The signals more than three are displayed with a two column layout and, if the number of total signal are odd, the last signal will be displayed at the last two bottom right frames of the layout, so that the two columns get the same length.
- WFDB Database AccessGreatMiner is designed as an universal tool processing data from different applications. WFDB database access gives a convenient means for ones interested in medical signal analysis to acquire data over Internet. People who work at other fields can also use this means to acquire data for experience working with this tool.
- Tips&TricksGreatMiner is designed under unique concept with many characters different from other tools in signal processing. Below tips and tricks can make it much easier for you to use GreatMiner efficiently and effectively.
Using the GreatMiner
- Conjunct Analysis This tutorial covers the procedures within GreatMiner that import multiple data or multi-parameter data and perform conjunct analysis.
The conjunct analysis in GreatMiner is defined as: The analysis at one or multiple signals according to the operation result of another signal. Conjunct analysis can be essential for some applications that treat multiple data from the same source; for example, the ICU signals come from the identical person that can be studied by conjunct analysis to discover relationships among the signals.
The conjunct analysis in GreatMiner is implemented by enable Bond Signal or by functionality.
Conjunct analysis by bonding signals
- Click to select Bond Signals at Tools > Options of Main Menu.
- Click on the Open File Button.
- Select a WFDB record, mgh003 in this example, six signals inside the record will be displayed in below multi-frame layout:
- Zoom the signal CVP_mgh003.txt by dragging a rectangle at the signal to include a cycle of the signal, all signals in other frames are also zoomed synchronously at the same time scale.
- Clicking Zoom out button at top right of one frame, the signals at all frame will be zoomed out.
Signal de-noising example Right-click at the signal CVP_mgh003.txt and select Common Operations -> De-noise option from Right-Clicking Menu, and drag stick of the horizontal slider at the bottom of selected signal, to modify de-noise depth until satisfied de-noise result achieved. All signals at other frames are also de-noised at the same de-noise depth.
You can perform Frequency Domain Analysis and Time-frequency Domain analysis at one signal of a multiple frame layout and all signals in other frames will be also performed the same operation.Conjunct analysis by functionality
Signal decompose example
GreatMiner gives a flexible way for signal decomposing. Because the wavelet transform decomposes a signal into different levels that are remained at the wavelet transform coefficient matrix, one row for one level of decompose result; we can achieve suitable decompose level by investigate into the wavelet transform coefficient matrix by below procedures:- Deselect Bond Signals option.
- Click on the Open File Button.
- Select a WFDB record, mgh003, six signals inside the record are displayed:
- Right click and select a signal, ECG lead II_mgh003.txt. From the right-clicking menu choose Wavelet Transform; the signal is wavelet transformed and the transform coefficient matrix is displayed at the second frame of below Four-Frame-Layout, in format of a matrix map image.
- Moving mouse at the wavelet coefficient map (second frame), the wavelet transform coefficient at a row (frequency) indicated by the cursor in the second frame will be displayed at the third frame. Moving mouse at the second frame and observing decompose result at the third frame until achieve satisfied decompose result (decompose at 0.045Hz for this example), right click mouse to remember the decompose level, all signals will be decomposed at 0.045Hz (equal the 90th level of a 100 level decompose in this example), and the results are displayed as below:
Cross-correlation analysis example
- Right click at a signal, ECG lead II_mgh003.txt at the multi-frame layout illustrated above to select this signal as a main signal, and all others are by default the target signals.
- From the right-clicking menu, select Common Operation -> Cross-correlation to calculate cross-correlation between the main signal and the target signals. The cross-correlation between main signal to all others are displayed together with relative target signals:
- Time-Frequency Domain AnalysisThis tutorial covers the Time-frequency Domain Analysis capabilities of GreatMiner. This includes the Wavelet Transform(WT) and Short Time Fourier Transform (STFT) spectral procedures in Multi-Frame-Layout or in Four-Frame-Layout.
Open multiple data files
- Click on the Open File Button.
- Select the six data files, leaving file filter being text files (*.txt).
Wavelet analysis – conjunct analysis
- Select Bond Signals in Main Menu > Tools > Options.
- Click on Wavelet Analysis in Time Frequency Domain Analysis from Main Menu >Tools or from Right-clicking Menu.
Wavelet analysis – individual analysis
- Deselect Bond Signals in Main Menu > Tools > Options.
- Click on Wavelet Analysis in Time Frequency Domain Analysis from Main Menu >Tools or from Right-clicking Menu. This will perform continuous wavelet transform at signal in selected frame, resulting at a Four-Frame-Layout.
- Dragging a rectangle at wavelet coefficient image will zoom the image inside the rectangle; clicking the zoom out button at the top right of the image will zoom out the image.
STFT analysis – conjunct analysis
- Select Bond Signals in Main Menu > Tools > Options.
- Click on STFT Analysis in Time Frequency Domain Analysis from Main Menu > Tools or from Right-clicking Menu.
- Deselect Bond Signals in Main Menu > Tools > Options.
- Click on STFT Analysis in Time Frequency Domain Analysis from Main Menu > Tools or from Right-clicking Menu. This will perform STFT at signal in selected frame, resulting at a Four-Frame-Layout.
- Dragging a rectangle at STFT coefficient image will zoom the image inside the rectangle; clicking the zoom out button at top right of the image will zoom out the image.
STFT analysis – individual analysis
- Fast Fourier Transform and Power Spectral Density This tutorial covers the Fourier Transform and Power Spectral Density capabilities of GreatMiner.
Fast Fourier transform and power spectral density calculation in conjunct analysis
- Click on the Open File Button.
- Select the six data files, leaving file filter being text files (*.txt).
- Ensure the Bond Signals option is selected.
- Select Fast Fourier Transform in Tools > Frequency Domain Analysis from Main Menu or in Frequency Domain Analysis from Right-clicking Menu:
- Select Tools > Time Domain Analysis from Main Menu or select time Domain Analysis from Right-clicking Menu to resume original signal.
- Select Power Spectral Density in Tools > Frequency Domain Analysis from Main Menu or in Frequency Domain Analysis from Right-clicking Menu:
Fast Fourier transform and power spectral density calculation in individual analysis
-
Fast Fourier Transform:
- Load signals into a multi-frame layout and select the Bond Signals option.
- Right-clicking a frame where the signal under analysis is located. A Right-clicking menu appears.
- Select Fast Fourier Transform in Frequency Domain Analysis from the Right-clicking Menu, Fourier transform spectral will be displayed at a single frame.
- Right-clicking and select Time Domain Analysis from Right-clicking Menu will resume original signals in the multi-frame layout.
- Load signals into a multi-frame layout and select the Bond Signals option.
- Right-clicking a frame where the signal under analysis is located. A Right-clicking menu appears.
- Select Power Spectral Density in Frequency Domain Analysis from Right-clicking Menu, Power Spectral Density will be displayed at a single frame.
- Right-clicking and select Time Domain Analysis from Right-clicking Menu will resume original signals in the multi-frame layout.
Power Spectral Density:
- Pattern TraceThis tutorial covers the pattern trace capabilities of GreatMiner for that instance where you want to trace patterns from a signal.
build a pattern
To trace a pattern from a signal, the first step is to build a pattern. GreatMiner allows you to build a pattern from currently displayed signals at a multi-frame layout.- Dragging a rectangle by mouse at a signal (ECG lead II_mgh003.txt), to include a portion of signal that you want to be a pattern. The signal inside the rectangle is zoomed and fills the frame.
- Repeat above step 1 until the zoomed signal is the required pattern.
- Right-clicking the zoomed frame, and select Build Pattern from Right-clicking menu to save zoomed signal as a pattern.
- Select Time Domain Analysis to resume the zoomed signal if you want to use the same signal as the target signal.
Pattern Trace
Trace Pattern by Time:- Select a target signal (ECG lead I_mgh003.txt) by right-clicking the frame where the target signal located.
- Select Trace Pattern from Common Operations from right-clicking menu, a sub-menu appears.
- From the sub-menu, select Trace Pattern by Time, portions of the target signal that are similar in timing to the pattern which is displayed blue and overlap the target signal, become green, with a horizontal slider indicating the similarity and a vertical slider reflecting the location of the pattern. More right the stick of the horizontal slider stays, higher the similarity. Dragging the stick of the vertical slider up will move the pattern right.
- Drag the stick of the horizontal slider right to set up a higher similarity value, resulting at less portions of the signal that are similar to the pattern, are found:
- Dragging the stick to the center of the vertical slider, the pattern is also moved to the center of the target signal.
- Select Trace Pattern from Common Operations from right-clicking menu, a sub-menu appears.
- From the sub-menu, select Trace Pattern by Amp, portions of the target signal similar to the pattern in amplitude become green, with a horizontal slider indicating the similarity and a vertical slider indicating the position of the pattern. In this example, there is no potion of the signal similar to pattern is found and below dialog appears:
- Click OK button to cancel the dialog.
- Select Trace Pattern from Common Operations from right-clicking menu, a sub-menu appears.
- From the sub-menu, select Trace pattern by time and amp, portions of the target signal similar to the pattern both in time and in amplitude become green, with a horizontal slider indicating the similarity and a vertical slider indicating the position of the pattern.
- In this example, no portion of the target signal similar to the pattern both in time and in amplitude is found and below dialog appears:
- Click OK button to cancel the dialog.
Tip:You can zoom the place where the pattern overlaps the target signal, to observe the accordance between the target signal and the pattern.
Trace Pattern by Amplitude:Trace Pattern by Time and Amplitude:
- Auto-Correlation and Cross-CorrelationThis tutorial covers the Auto-Correlation and Cross-Correlation capabilities of GreatMiner for extracting relationship among signals.
Auto-correlation
- Click on the Open File Button.
- Select the six data files, leaving file filter being text files (*.txt).
- Select the signal ART_mgh003.txt by right-clicking the top left frame, a right-clicking menu appears.
- Select Auto-correlation in Common operation from Right-clicking Menu to calculate auto-correlation of selected signal, the correlation results are displayed and colored green, together with original signal:
- Right-clicking the frame where the auto-correlation was performed, select Common operation > Switch single-multi frame from right-clicking menu to observe auto-correlation results in single frame. This step is optional.
- Click on the Open File Button.
- Select the six data files, leaving file filter being text files (*.txt).
- Select the signal ART_mgh003.txt by right-clicking the top left frame, a right-clicking menu appears. The selected signal will be used as a main signal to calculate cross-correlation with all signals in this multi-frame layout.
- Select Cross-correlation in Common operations from Right-clicking Menu to calculate auto-correlation of selected signal, the main signal is performed Cross-correlation with all other signals and the results are displayed in green together with target signals:
Cross-Correlation
Key Information:
Note that the main signal also calculate cross-correlation with itself, in this case, the results equal the results of auto-correlation.
Key Information:
Note that the correlation results are not normalized and with lap 0 at the center of X axis. - Noise ReductionThis tutorial covers the denoise capabilities of GreatMiner. Noise can be removed from data using Wavelet decomposition and reconstruction techniques with the procedure being controlled manually or automatically.
Denoise manually
- Click on the Open File Button and load multiple signals:
- Select the signal at bottom frame by right-clicking, and click De-noise in Common operation from right-clicking menu. the signal is denoised at a default threshold (80% of maximum threshold) and a horizontal slider indicating the threshold level:
- Drag stick of the horizontal slider to changed denoise threshold until satisfied result reached.
- Click on the Open File Button and load multiple signals:
- Select the signal at bottom frame by right-clicking, and click Auto-de-noise in Common operation from right-clicking menu. the signal is denoised according to a Auto-Denoise Algorithm:
- The signal at the top frame is also denoised equally, if Bond Signals option is selected:
The signal at the top frame is also denoised equally, if Bond Signals option is selected:
The signal at the top frame is also denoised equally, if Bond Signals option is selected:
Denoise automatically
- Signal DecomposeThis tutorial covers the procedures within GreatMiner that decompose a signal by wavelet transform.
In this example, below signals are loaded into a multi-frame layout:
Decompose a signal at a predefined level
- Select Tools > Signal Decompose from Main Menu
- Click Set Decompose Level. A Input dialog appears, asking you inputting decompose level.
- Input decompose level 100 and click OK to setup the decompose level.
- Select Tools > Signal Decompose from Main Menu again and click Decompose. All signals at the layout are decomposed at level 100:
- Deselect Bond Signals in Tools > Options of Main Menu.
- Click on Wavelet Analysis in Time Frequency Domain Analysis from Main Menu > Tools or from Right-clicking Menu, resulting at a Four-Frame-Layout.
- Move mouse at coefficient image in the 2th frame and observe the decomposed component at the 3th frame until required de-composition reached. Right-clicking mouse.
- Select the Get decompose level option. All signals at previous multi-frame layout are decomposed into current level indicated by cursor position in above step 3 and the decompose results take place of the signals before decompose in previous multi-frame layout.
- Deselect Bond Signals in Tools > Options of Main Menu.
- Select the signal ECG lead II_mgh003.txt by right-clicking, and select Wavelet Analysis in Time Frequency Domain Analysis, resulting at a Four-Frame-Layout indicating Wavelet Transform results.
- Right-clicking at wavelet transform coefficient image and select the Search decompose level option. the program will search decompose level by search decompose level algorithm and decompose all signals at previous multi-frame layout. The previous multi-frame layout is displayed again with the decompose results taking place of the signals before the decompose.
Tip
If no decompose level was set up, a default level 80 will be used.
Key Information
Note that all the signals at a multi-frame layout will be decomposed no matter if the Bond Signals was selected or not.
Decompose a signal by set decompose level through Continuous Wavelet Transform
Decompose with the decompose level searched manually:
Decompose with the decompose level searched automatically:
Key Information
Note that selecting different signal to perform wavelet transform and search decompose level may result at different decompose level.
Tip
You can always decompose the signals in current layout no mater it is the signal original or the signal after processing. The default decompose level or the level set by last set-level-operation (if there is) will be used for decompose.
Learning from examples