UCLA Phonetics Lab
The Phonetics Lab owns a Carstens system (AG100) for electromagnetic articulography (EMA). Currently we have 10 receiving pellets, which can be placed midsagitally inside the mouth or on the face to measure movements during speech production.
| Development of this site has been supported by grants from the
UCLA Academic Senate Committee on Research, and NSF grants SBR-0001716 and
IIS-9996088, for which we are grateful. However, the
information found in this website is based upon our own experience and tips from other people. It is possible that this website contains inappropriate or misleading information, for which funders should not be blamed!
STEP 0: Sterilization and DisinfectionSterilize and disinfect materials (sensor coils and tweezers) by using Cavicide (can be bought from Amazon). Immerse sensors and tweezers following instructions for STERILIZATION
Less than sterilization is possible, but not usually used here: immersion in glutaraldehyde for a minimum of 20 minutes for HIGH LEVEL DISINFECTION; immersion in ethyl alcohol (75 ~90%) for 20 minutes followed by wiping with Povidone-Iodine can be used for disinfection only (not for sterilization).
- Relevant Definitions (source: Classes Of Disinfectants And Their Uses)
- Antiseptic: An agent that can be used to reduce the microbial population found on skin.
- Disinfectant: An agent that will destroy many of the disease-causing microorganisms present on the surface of an inanimateobject.
- Sterilant: An agent that destroys all microbial organisms including heat-resistant bacterial spores. Sterilization can be achieved by boiling, autoclaving or exposure to toxic chemicals. Solutions that contain chlorine or glutaraldehyde are frequently used as chemical sterilants.
STEP 1: Warming-UpThe system must be turned on at least two hours before the actual experiment, in order to allow the coils to warm up and reach a stable temperature. Once the computer and the EMA system are on, it is necessary to start the Art program for the warm-up procedure to begin (after Art starts, you can exit the program and use the computer for other purposes). See the Carstens website and the UCLA homemade manual for details on Art.
STEP 2: Coating SensorsDip the sensors into Plasty-Late (modelers plastic) to coat them. This step protects the sensors by facilitating adhesive removal after the session and also helps keep them clean. (It is also fine to dip the wire about 1-3 inches into Plasty-Late). The solution is milky when wet but transparent when dry.
STEP 3: Attaching Silk Cloth(Optional but Recommended) Attach the sensors to small circles or squares of silk cloth right after dipping them in the plastic coating (so the cloth sticks to the sensor). (See Figure (3) below.) As can be seen in Figure (4), the cloth expands the surface of the sensor coil that is attached to the tongue and consequently helps preventing the sensors from falling off. This step is only recommended for coils that will be attached to the tongue.
STEP 4: Drying the Plastic CoatingIt is important to leave the plastic coating to dry for at least 2 hours (the longer the better - as a general rule, we leave the coating dry for one night), at which point the sensors can be calibrated in the magnetic field. Insufficient drying may cause the coating to be broken more easily, which results in the sensors' falling off.
[Back to Top]
STEP 5: Positioning the helmetThe helmet which generates the field is positioned around the subject's head. Since the helmet is rather heavy, it is suspended from a pulley, in order to alleviate some of the weight.
- We have found that by placing the sensors before putting on the helmet, the subject has some time to adjust to having the sensors attached, and also does not have to sit so long with the heavy helmet on.
- However, a drawback may be that one cannot check if the attached sensor coils are correctly detected in the magnetic field during the attachment session.
- Experienced experimenters may find it more efficient to place the sensors before putting on the helmet, especially when the speech material to be recorded is relatively long. In such a case, examine carefully whether sensors responds to the system constantly as the lead at the coil base out is gently twisted around. This is usually done when sensors are calibrated.
[Back to Top]
STEP 6: Placing the sensors
A. Calibration ProceduresIt is essential to perform a calibration of the system before each experiment.
Please go to our UCLA Homemade EMA Manual for detailed Calibration Procedures.
B. Materials for AttachmentIn order to place the sensors, several materials should be prepared.
- For internal use (inside the mouth): dental adhesive, formerly Cyano-Veneer® Fast , now a PeriAcryl cyanoacrylate dental formulation tissue adhesive from Gluestitch.com, but we note that Haskins uses Ketac-Bond cavity liner
- For external use (outside the mouth): skin adhesive, formerly Skin-Bond® by Smith and Nephew, now a liquid bonding cement ordered from Amazon, but we note that Haskins uses Iso-dent from Ellman International
- Reversal Tweezers
- The reversal tweezers help in holding the sensor coil with precision and not dropping it. Reversal tweezers can be ordered from a surgical supplies store.
- Medical Tape
- To affix sensor coils to the face and lips (instead of using Skin-bond).
- To tape the leads to the cheek so that they don't get pulled off.
- Gauze and/or Cotton Pad
- To wipe the tongue in order to dry the tongue surface (we found that using cotton pads is better than using gauze).
- Cosmetic pencil (not in figure)
- To put a dot on the point of the tongue to which the sensor coil is attached.
- Hair-Dryer (not in figure)
- To dry the tongue tissue thoroughly. It is important to use a dryer with a "cold lock" option.
- Tissues (not in figure)
See UCLA Homemade Manual for Performing an Actual Measurement Sweep.
[Back to Top]
Processing the data [using Tailor]
- Before using Emalyse to measure position data, several processing steps are needed using the Tailor software. Please refer to Tailor manual.
- Filtering (see also the Tailor manual p.7-9)
- Filter_40: this filter is the strongest for smoothing. It is recommended by Carstens to use this filter for fixed reference points. The cut-off frequency is at sample frequency/25. At a sample frequency of 1000 Hz, the cut-off frequency if 40 Hz and the filter offers a 60 dB depression at 90 Hz. (The cut-off frequency, therefore, depends on the sample frequency.) USE THIS FOR SENSORS AT TWO REFERENCE POINTS (UPPER GUMLINE, NOSE BRIDGE.)
- Filter_100: The cut-off frequency is at sample frequency/10. As a sample frequency of 1000 Hz, the cut-off frequency is 100 Hz and the filter offers a 60 dB depression at 150 Hz. USE THIS FOR MOST SENSORS.
- Filter_160: This filter is for weak smoothing of movement signals. The cut-off frequency is at sample frequency/6.25. At a sample frequency of 1000 Hz, the cut-off frequency is 160 Hz and the filter offers a 60 dB depression at 210 Hz.
Dynamic correction for head movement
- The dynamic correction procedure modifies the co-ordinate system individually, sample by sample, in such a way that both reference sensors stay still. This step eliminates the errors caused by helmet movement during the recording session. (from p.9 in the Tailor manual)
Rotation to the occlusal plane
Restoring 12 bit acoustic data [using Multi-CV]
- First, determine carefully by what angle the data needs rotating. This can be determined by using x'(distance between two x points) and y' (distance between two y points) values of two points on the bite plate. The angle is the arctangent of y'/x' (theta = tan-1 y'/x').
- Enter the angle determined in the box provided. This process rotate the entire data.
- The speech processors generate, store and process the acoustic data in the compressed 4-bit ADPCM format. Emalyse can read only this compressed format. Thus, the acoustic data in Emalyse should not be used for any serious acoustic measurement.
- However, the compressed acoustic data can be 'uncompressed' to the 12-BIT PCM format which can be used for serious acoustic measures. Multi-CV allows this conversion. First, open a data file in the Multi-CV, mark "sound-wav" function, choose the sweep numbers you are interested in. Let it run. The 12 bit acoustic data (.wav) will be stored in the folder "Convert" in the same directory in which your original data are.
Analyzing the data
- we use the EMALYSE software to measure position data.
- Here is an example display from the Emalyse program, showing two Croatian sentences (Kuda Bibi pada, "Where does Bibi fall?" and Kuda Mimi pada, "Where does Mimi fall?"):
*An animation of the articulators in motion, created from EMA data, is also available on the Demos and Illustrations page.
Last updated: content 02/18/00 (by Taehong Cho), links Oct. 2017 (Pat Keating)