AG100
The Tailor Program
How to operate the Tailor program
The Tailor Program
General Information
With the Tailor program you are able to modify measurement data acquired by the Articulograph AG100. These modifications will optimize the investigation conditions afterwards.
Information that is not necessary for data analysis can be removed. It is also possible to create data that was not recorded in real.
The operating program Art.exe divides an investigation into a sequence of sweeps. A coherent record is named a "sweep".
The Tailor program can act on one sweep at a time or on all sweeps from the whole data set.
The Tailor program is a part of the Windows program package for working on the Articulograph AG100 investigation data. The Emalyse - Tailor - MultiCV programs complement each other.
While the Emalyse and MultiCV leave the measured information completely unchanged, it is the Purpose of the Tailor program to process and modify the measured information.
Normally, the Tailor program will overwrite original measurement data. Please use the Save as.. function to store the modified data in an other location and to leave the original data unchanged.
The purpose of the Tailor program is to do all necessary data modifications, while the MultiCV and Emalyse programs leave the investigation data itself unchanged.
The Tailor program has four function groups. Two of them affect the whole data set and the other two only affect one sweep or a part of one.
Global Edit
The Global Edit functions modify the co-ordinate system of the movement data, smooth the movement curves, etc..
One special "global edit" function is to run a script file that was created by the Emalyse program. This way it is possible to use the comfortable analysing and displaying functions from the Emalyse program to mark all timing points to cut off areas, to create new sweeps and to build average movement curves.
Sort & Concatenate
The Sort and concatenate sweeps function allows you to change the variety of sweeps inside the data set. There is an option to edit a sweep's comment; new sweeps are created by concatenate several sweeps.
Sweep Edit
The Sweep Edit function allows the modification of a single sweep at a time.
A sweep can be shortened by cutting a part off at the beginning or at the end. The sound signal can be forced to complete silence for the whole sweep or for time segments in it.
Sweep Composer
The Compose new sweeps function looks like the sweep edit function, but leaves the sweeps unchanged. It creates new sweeps in two ways and appends them at the end of the data set. The new sweeps are either a portion from an existing sweep or the average movement from several portions.
The main window also offers a speed bar with function buttons.
Result from data processing with Tailor
The modified data set has no differences compared to a data set that is recorded by the Art program. The data format and the type of files do not change. Only the *.prt file has an additional line: "Tailor=yes". Of course, from the users point of view, there are differences caused by the purposes of the Tailor program.
The modified date is stored in the selected folder and has overwritten the previous data.
The program also offers "save as.." options to redirect the destination of the modified data set.
With the Save 
Button or on leaving the Tailor program the result of
the Tailor session will be stored in the original location and the previous
data will be overwritten.
With the Save As.. 
button you can store the modified data set in another
folder and leave the original data unchanged.
Every study creates 3 files with organisational data and additional 3 files for each sweep. The sweep number is coded in the file extension. The first character of the file extension defines the type of the data (0= co-ordinates; T= tilt factor; A= acoustic) and the last two characters represent the sweep number.
If you name a study, for instance "TEST", you get the following files:
Files with organisational data
The following files are only created once for the study.
TEST.PRT
The protocol file "TEST.PRT" is a normal text file and contains the system parameters you have adjusted, the protocol data and the commentary to the individual sweeps. You can type, print, or edit this file.
TEST.CFG
This file contains the adjusted system parameters and calibration information in a binary format. This and the following file are used by the evaluation programs. For special analysis you need programmer information. The Type declaration is shown in the chapter "Detailed programming information".
TEST.TIM
The timing file contains one data record for each sweep. The record is named Tim_TYPE. It contains the sweep number, the number of measurement points and the starting time in hours, minutes, seconds and hundredths of seconds. The Type declaration is shown in the chapter "Detailed programming information".
Files for each sweep
The following three types of data files are created for each sweep in a study. The last two characters of the filename extension represent the sweep number.
TEST.001 .. TEST.099
The movement data from the first sweep is stored in "TEST.001". The XY-co-ordinates are in units of 0.01 mm as integer values.
TEST.T01 .. TEST.T99
For every XY-co-ordinate stored in TEST.001 there is a tilt factor stored in the file TEST.T01.
TEST.A01 .. TEST.A99
Each sweep's acoustic speech information is stored in these files in a compressed format.
Additional files, when more than 5 channels are used
TEST.101 .. TEST.199
The movement data of the channels 6 - 10 are stored in "TEST.101", in the same manner as the data of channel 1 - 5 are stored in TEST.101".
TEST.U01 .. TEST.U99
For every XY-co-ordinate stored in TEST.101 there is a tilt factor stored in the file TEST.U01.
Additional files, when more than 10 channels are used
TEST.201 .. TEST.299
The movement data of the first sweeps is stored in "TEST.201". The XY-co-ordinates are in units of 0.01 mm as integer values.
TEST.V01 .. TEST.V99
For every XY-co-ordinate stored in TEST.201 there is a tilt factor stored in the file TEST.V01.
Cut, sort and concatenate sweeps
With these functions you can build a data set, tailored to the requirements for best data analysis. The cutting of sweeps has the same effect as as if the recording was started later and stopped earlier.
The sorting of sweeps has the same effect as as if the sweeps were recorded in a different order.
Concatenating sweeps also allow the construction of utterance sequences that could not be recorded.
Shift, rotate and mirror the co-ordinate system
The co-ordinate system for the Articulograph AG100 data is determined by the helmet position during the investigation. These functions modify the data in a way as if the helmet had been at another position during the investigation. This way you can get precise helmet positions that can not be achieved in a real investigation environment.
The purpose is to have the bite plane parallel to the X axis and significant (reference) points at the same location for all investigations.
Smooth and compress the data by digital filtering
These functions are to minimise disturbances that are overlaid by the measurement electronic. The data looks as recorded with an ideal measuring system.
Compression is done with help from a digital Filter (digital Filter). The number of samples is reduced to half and each sample seems to be measured twice.
With help from two reference sensors, the co-ordinate system for each sample is individually newly positioned. The positioning is done in a way that the reference sensors stay at a fixed position. This eliminates unintentional helmet movements relative to the head.
Several segments with the same utterance can be put into a sweep that represents the average speech movement.
Data from sensors that were used for references can be removed from the data set after the validation procedure.
The sound signal from a whole sweep or from a part of the sweep can be set to silence.
The Tailor program offers the smoothing of the sensor signals with a digital filter. There are three different types of filters used in the Global edit-IDH_glMain an in the sweep edit-IDH_swMain window. A fourth filter type is internally used for the compress dataIDH_gl-downsampel function.
All filters are a Hamming windowed finite impulse response type (FIR). They have an operator length of 101 (201 for the data compression filter).
To minimize the errors at the beginning and the end of each smoothed curve, the curves are extrapolated before the filtering procedure. There is no error if you run the filter first and then cut off the beginning and the end of the sweep. If you cut more than 50 samples off, there is no error left.
The selectable filters differ in their strength of smoothing. The cut-off frequency depends on the sample frequency used while recording the data.
This filter is the strongest for smoothing. It is recommended to use this filter for fixed reference positions. The cut-off frequency is at Sample frequency /25.
At a sample frequency of 1000Hz, the cut-off frequency is 40Hz and the filter offers a 60 dB depression at 90 Hz.
This Filter is for strong smoothing of movement signals. The cut-off frequency is at Sample frequency /10.
At a sample frequency of 1000Hz, the cut-off frequency is 100Hz and the filter offers a 60 dB depression at 150 Hz.
This Filter is for weak smoothing of movement signals. The cut-off frequency is at Sample frequency /6.25.
At a sample frequency of 1000Hz, the cut-off frequency is 160Hz and the filter offers a 60 dB depression at 210 Hz.
This Filter is for data compression. The cut-off frequency is at Sample frequency /5.
At a sample frequency of 1000Hz, the cut-off frequency is 200Hz and the filter offers a 60 dB depression at 225 Hz.
The normally used sampling frequency is lower than 1000Hz. The values for the cut-off frequency and 60 dB depression change at the same scale as the sampling frequency.
If you run the filter function on already filtered data, the effective cut-off frequency is slightly reduced and the depression is much better.
Example Data processing procedure
The following suggestions/instructions describe an example procedure. Thus at least two birds are killed with one stone: first, you get to know how and in which order the pre-processing of your data should be organized best. Second, you get familiar with the Tailor functions. After working yourself through the following instructions and suggestions, it will be easy for you to find your own most effective procedure, which match your individual demands.
The sweep edit and sweep composer functions are useful for getting a first insight of how to modify sweeps. However, to process large data sets, it is much more effective to place markers and segments by means of the Emalyse program. Afterwards the Tailor will automatically do the sweep modifications by using the "Run Emalyse Script" function.
Run Tailor for the first step of data treatment:
down sample
Assuming your data set was recorded at 500Hz. In a first step, you can select the down sample function. Now the data set looks as recorded at a sample rate of 250Hz, but it exhibits a much higher quality, because each sample contains the information of two samples.
smooth
The best time for smoothing is after down sampling and before further pre-processing, because at the beginning and the end of each curve the data is extrapolated by 50 samples. Later, if you cut 50 samples or more, you will still get a 100% correct data.
Select Filter_40 for both reference sensors. The filter affects with a cut-off frequency of 10Hz. This strong smoothing is recommended, because we do not expect any movements at the sensors on the upper incisor or at the nose.
The two sensors on the lips are smoothed best with Filter_160, because we have recorded high frequency lip flutter. The filter affects with the cut-off frequency at 40Hz.
The remaining channels are all smoothed with Filter_100a_Filter, which has the cut-off frequency at 25Hz.
dynamic correction
The dynamic correction procedure modifies the co-ordinate system individually, sample by sample, in such a way that both reference sensors stay in place. This step eliminates the errors caused by helmet movement during the recording session.
mirror
After recording with the Art program, the subject always looks to the left side of the screen. If you want the subject to look to the right side, please select the mirror function.
save as..
Use the Save as.. command to store the data set in a new folder and to leave the originally measured raw data unchanged.
Save Log File
Save the history list in a log file to document how your raw data set was processed up to now. Then close the Tailor program.
Run Emalyse
Now run Emalyse to extract the part of the data you are interested in, for example target words or sentences. This step is useful for data reduction.
Use the information from the sensors on the bite plane to find the angle between the bite plane and the X-axis. With this angle the Tailor program can rotateIDH_gl-shift-rot the co-ordinate system to make the subjects bite plane (or other reference lines) parallel to the x-axis.
Find the positions where data of interest begin and end in each sweep. Position cut markers at these time points (see note 1). For example, if you recorded utterances like "I said ta, ta, ta again", you only need the target syllables "ta, ta, ta" for analyses.
Run Tailor for the second step of data treatment:
While the program is closed, the Emalyse program will store all actual marker and segment information to disk. The Tailor uses this information as "Emalyse Script".
Start the function "Run Emalyse Script"
This procedure will automatically extract the data according to the markers you have created in the Emalyse program.
Rotate
Use the shift and rotateIDH_gl-shift-rot function to rotate the co-ordinate system in a way that the subjects bite plane is parallel to the X-axis.
Run Emalyse for more detailed analysis
The sweeps are now shortened and you need less zoom to find significant points in the displayed curves. As you will notice it simplifies the positioning of markers and segments.
Now you can either select segments for creating new sweeps and assembling average movement curves in all sweeps (see note 2) or extract data at significant points, for example minima or maxima in the time signal, articulator positions at an acoustically defined moment etc. (see Emalyse for details).
For a better comparison between subjects or multiple recordings of the same subject, it is very useful to keep reference sensors at the upper incisor in the same point. Find out the distance between the actual reference point and the desired position and keep these numerical values for the next Tailor run.
Run Tailor for the third step of data treatment:
Close the Emalyse program to store the marker and segment information for Tailor. Start the Tailor.
Shift
Move the co-ordinate system by means of the shift and rotate function to position the reference sensor at the desired place, so it is comparable with other studies.
Erase channels
Data of reference sensors are not necessary for further processing. Use the erase channel function to delete these channels. It improves the clearness of the display of the remaining channels.
save as
Save the complete post processed data at a new location. All following steps can concentrate on analysing the data to evaluate parameters from the movement and acoustic data (see note 3).
Save the Log File
Save the history list in a log file to document how your raw data set was modified up to now.
Start the function "Run Emalyse Script"
This function will create new sweeps according to the markers you have positioned in the Emalyse run. Apart from that, a sweep with the average movement curves is created.
Sort new data sets
Assuming that you want to analyse the current investigation, for example under three different aspects, please use the sort & concatenate window three times to build three different data sets. This is done by selecting all desired sweeps and using the "save as.." function to store the data in a different folder. Then open the previous data set again, select all sweeps you need for the second aspect and repeat the steps until you obtained the appropriate data set for each analysis aspect.
Concatenate sweeps
With the concatenate function you can now build new sweeps with a sequence of all first ta's, all center ta's and all last ta's. You can compare their movement course with the average movement built over all ta's.
One data set may contain all sweeps to analyse the utterance of "ta". Besides the original sweeps with "ta, ta, ta", you now have newly created sweeps with single ta's. They are sorted according to their original position during the utterance.
Run MultiCV for exporting the data
Convert into ASCII text format for calculation programs
At this point, the arranged data are ready prepared and concentrated to build the final result of the investigation. With the MultiCV program you can convert a complete data set or a single sweep into ASCII data format.
Adjusted the format to the requirements of your calculation- or data base programs.
Convert to AutoCad for graphic presentation
To build graphic presentation from average movements or other movement sequences with AutoCad, convert the desired sweeps into AutoCad format.
Note 1:
It is also possible to cut sweeps with the sweep edit function in the Tailor program.
Note 2:
It is also possible to create new sweeps and ensemble average movement curves with the sweep composer function in the Tailor program.
Note 3:
This is an example procedure of data treatment to give you an idea of how all the functions will work. It is also possible to use the functions in a different order and it is not necessary to use all functions. Please use this as a suggestion or an example and create your own procedure according to your special needs.
How to operate the Tailor program
The Tailor delivery disk contains one folder named "Tailor". Please copy the complete folder to your hard disk. Open the new folder on you hard disk and use the drag and drop function to bring the "Tailor.exe" file onto your desk top.
The program symbol (icon) for the Tailor program will appear on
your desktop now. Please double click this symbol to start the Tailor program.
Click the Help button to start the Tailor Help system. It may be, that the program can not find the File "Tail_h.hlp" and displays a message. In this case, please make sure that it is in the Tailor folder and select this file.
The Tailor program is a 32 Bit program and requires at least Windows 95 or Windows NT.
The free disk space has to be more than three times the used capacity of the current data set plus 12 mega bytes.
The following files are necessary to run the Tailor:
Tailor.exe
Cursor.dll
Tail_h.hlp
Tail_h.cnt
The Tailor program works on one data set at a time. The data set must be a set of files in one folder (sub directory) generated by the Art program.
Tailor is also able to use markers and segments, defined by Emalyse
The display function can snap to Emalyse marker to adjust its own markers in the predefined positions.
The Run Emalyse script file function uses the marker and segment information to automatically do all modifications inside the whole data set.
The Tailor and its functions are described within the Windows help system. There are several ways to activate the help system:
The 
button in the main window starts the Tailor help with the
index page. Other windows also have their own 
button to lead directly
to the specific description.
The Question mark symbol is in the upper right corner, the right mouse button and the F1 key offer context sensitive help information. A help page with help information to the currently selected control element will appear.
How to use the Tailor Functions
A double
click on the Tailor icon starts the tailor program and shows the main menu:
The main window appears with the most buttons in the colour grey. Only four buttons are not grey and their functions will work. Please click the button to load the desired data set. The Tailor program needs all files that are generated by the "Art" program during the recording session.
A dialog to select a file will appear. Only Ag100 data-files files with the extension *.tim are offered. Please change the folder until the filename for your desired data set appears. A double click on this filename loads the data.
All buttons except "save data" are coloured now and will work. The white panel shows the actual loaded data set.
What to do first
Save as
The Tailor program will store its result in the same folder as you selected with the "load data set". The "Save as" function will redirect the data storage to another folder.
If you want to keep the raw data unchanged, please use the "Save as" function to select a different folder. You must use the "Save as" function first, if your data are on a CD Rom or another 'read only' device.
Improve the data quality
Generally it is not very important in which order you perform the single steps for data processing. The only things are down sampling and smoothing. These procedures have less functionality at the beginning and the end of the curves. If these procedures are done before cutting the sweeps, you can keep a 100% correct data.
On the other hand, it is recommended to do all necessary things that generally improve the data quality first:
Prepare the data for evaluation
After the general data treatment, you can start with the preparation to make the data suitable for your desired evaluation. At this point, it depends on your requirements, in which order you do the steps for the data preparation. It is only recommended to cut off unused data first. The shortened sweeps need less zoom to find significant points in the displayed curves.
Normally the following steps are repeated several times in an independent order. The automatic data treatment functions are also available as single functions in the Sweep Edit and Sweep Composer Windows.
On Start the OK button is grey and all other function buttons are coloured and available for selection. The Global Edit does not start the selected functions immediately. First a description appears on the list. The execution of all selected functions in the list is started with the OK button.
What you have to pay attention to
Shift & rotate
If you use the shift and rotate function in the same run, please note that the rotate operation will also shift the xy positions. The best way is to rotate the co-ordinate system first and then calculate the desired distances for shifting in x- and y- direction.
Downsample and smoothing
These two functions may restrict the correct representation of high frequent movements. The down sampling sets the effective sample frequency to half of its original value. The sample frequency is the only parameter to determine the strength of each filter. The resulting cut-off frequency is shown for each filter in the smoothing dialog:
The "Filter_100" has a cut-off frequency of 100Hz if the sample rate is 1000Hz. The drawing shows a cut-off frequency of 40,0 Hz, because the actual data were sampled at 400Hz.
Please note that fast movements with their frequency equal to the cut-off frequency are damped to 70% of their original amplitude. Select a filter with a cut-off frequency that is more than 30% higher than the maximum frequency you expect.
Dynamic position correction
The calculation for dynamic position correction must have the information from two fixed reference channels. The according sensors from these channels must not have any movement relative to the subject's head.
In case of unintentional movement at the reference sensors, half of this movement amplitude is modulated on all movement curves.
Single functions
There are six function buttons in the section 'functions to modify all sweeps in the actual data set'. For more detailed information on each function, please click them with the right mouse button and select 'Online Help'.
Once you have selected one of the single functions, the according button and the 'Run Emalyse Script' button become grey. The OK button is now coloured to indicate that you can start the execution. The other single functions stay also valid for selection.
Automatic data treatment
The 'Run Emalyse Script' button selects the automatic
data treatment function. This line appears in the List:
'Automatic sweep modification by Emalyse script'
and all single function buttons become grey.
Start execution with the OK button
In case 'Automatic sweep modification by Emalyse script' was selected, all sweeps are searched for markers and segments that were created with help of the Emalyse program. The data treatment is performed sweep by sweep according to the marker respective segment type. For each sweep one line appears with the sweep number and the performed actions. For unmodified sweeps, only its sweep number appears.
If single functions were selected, the list stays unchanged.
In both cases, the progress of data treatment is displayed in extra windows. The program is busy until the main window appears.
A detailed list of the performed functions is available with the info(i) button.
Save
The modified data are stored in a temporary folder and all further actions will affect them. You are free to update the actual folder now or at any time, before leaving the Tailor program.
The sort & concatenate window will not execute the data modifications unless you click the OK button. All the time before, the actions are performed with the sweep comments acting for the sweeps.
The window has two lists with sweep comments. The left list 'actual data set' represents the actual data set and can not be modified. The purpose of the Sort & Concatenate window is to build the 'new data set' list with help of the window functions. The data modification is done according to the 'new data set' list. The sweep at the top of the list becomes the first sweep with the new sweep number 1.
Sort the sweeps in a new order
There are several methods to line up the sweeps in a new order. Use the mouse together with the shift or cntrl key to select the desired sweeps in the 'new data set' list. Copy the selected lines with the 'drag and drop' mouse function or with the right red arrow button to the 'new data set' list. The button will append the lines at the end.
The red up/down buttons moves selected lines inside the 'new data set' list.
Concatenate sweeps
This function builds a new sweep by concatenating all selected sweeps from the 'actual data set' list. The comment of the new sweep is appended to the 'new data set' list and consists of all the sweep numbers of the concatenated sweeps. An '&' sign is inserted between the sweep numbers.
Delete sweeps
There is no explicit function to delete sweeps, but all sweeps that do not appear in the 'new data set' list are removed from the data set.
Copy sweeps
There is no explicit function for copying sweeps, but each sweeps is generated as often in the new data set as its comment is in the 'new data set' list.
Save
The modified data are stored in a temporary folder and all further actions will affect them. You are free to update the actual folder now or any time, before leaving the Tailor program.
On start, the Sweep Edit copies the actual data into an extra working folder (name: Temp) and all modifications affect on this data. With the OK button, these data are copied to the working folder and become the actual data set.
The sweep editor offers basically the same functions as the automatic data processing with the 'Run Emalyse Script' function in Global Edit, but has some advantages.
While the all Global Edit functions always affect the whole data set, is it possible with the Sweep Edit to limit the functions to selected data areas. This selection is possible by positioning the adjustable Sweep Edit Marker at the desired time points inside a sweep. You can position the Marker with help of the scroll bars, or let them jump to the next Emalyse Marker.
This offers the possibility for example to smooth only one part from a sweep or set the sound signal to silence in another part.
The sweep is shortened at the beginning and the end according to the adjusted marker's positions.
Set the sound signal to silence
The sound signal in the area between the markers is set to zero.
Smooth the movement curves
The movement curves in the area between the markers are digitally filtered. The same dialog as in the Global Edit function appears for selecting channel and filter type.
Save
The modified data are stored in a temporary folder and all further actions will affect them. You are free to update the actual folder now or any time before leaving the Tailor program.
The Sweep Composer is similar to the Sweep Editor in sweep handling and display, but has two different functions.
Create a new sweep
A new sweep is created and joined to the actual data set. The length is determined by the positions of the markers.
Add the selected data to the average movement curves
The xy values from the movement curves are added to the common curves for building the average movement curves. The average movement curves are calculated when leaving the Sweep Composer with the OK button.
Please note, if you do not use the Emalyse Marker for this function, you must make sure that all data areas have the same length to build the correct average. It is also important to find the right reference position with the same distance to the begin- and end point.
This is the place to add extra tools in further Tailor revisions. The actual revision 1.3 offers the functions to save and clear the log file.
There is no explicit undo function, but you have several points to return to a previous state of data modifications. All four main functions have a Cancel button.
With Save as function, you can leave the actual data set at the status from the program start respectively the last Save function call.
You can undo the whole data modifications from complete Tailor session if you do not call Save or Save as during the program session and confirm the warning message on leaving the Tailor.