Phase coherent ultrasonic tracking
More actions
Phase coherent ultrasonic tracking is a type of ultrasonic tracking that measures the phase of the signal and compares it with a reference signal to determine the translation of a moving point. It was one of the first types of 3D tracking used for a head-mounted display, and was used in Ivan Sutherland's 3D display.
Instead of timing short pulses of sound, a phase coherent tracker drives its transmitters with a continuous ultrasonic tone and measures how far the received wave has shifted in phase. A shift of 360 degrees corresponds to a change in distance of one wavelength, so the method gives fine, continuous distance changes but no absolute distance on its own.[1][2] The best-documented example is the continuous wave ultrasonic head position sensor built at the MIT Lincoln Laboratory for Ivan Sutherland's head-mounted display experiments, described in Sutherland's 1968 paper "A head-mounted three dimensional display".[3] The method only measures relative motion and is easily disturbed by sound reflecting off walls. Commercial acoustic trackers used pulsed time-of-flight ranging instead, and development of phase coherent trackers was abandoned.[2][1]
How it works
Phase coherent methods work like this: the difference between two successive measurements of phase allows the distance change since the last measurement to be computed.[1] The translational movement of an ultrasonic emitter results in the receiver sensing a different phase. The receiver compares the sensed signal with a reference signal to determine the phase.[1] It is a relative system.
Welch and Foxlin describe Sutherland's system as a "continuous carrier-phase" tracker that used a continuous-wave source and determined range by measuring the phase shift between the transmitted signal and the signal detected at a microphone.[2] In a 1993 overview of tracking devices written for the student-produced Encyclopedia of Virtual Environments, Gregory Baratoff and Scott Blanksteen explain the geometry: the phase of a sound is its position on the wave, measured in degrees, and 360 degrees equals one wavelength of difference. As long as the target moves less than one wavelength between updates, the system can follow it, and using several transmitters allows orientation to be determined as well as position.[4]
Sound can only give the distance between two points, so acoustic trackers of either kind use several emitters and several receivers (typically three of each) with known geometry, and calculate position and orientation from the resulting set of distances.[1] A University of South Australia lecture on VR tracking summarizes the phase coherent approach as a continuous signal with no pulse, many transmitters on different frequencies, and phase differences between sent and received signals that give the change in distance continuously and without latency.[5]
To turn relative changes into an absolute distance, the tracker must know the starting distance and keep count of the whole cycles that have accumulated since then.[2] Any mistake in that starting value remains as a fixed offset in every later reading.[3]
Sutherland's ultrasonic head position sensor
Sutherland's display could use either of two head position sensors, one mechanical and one ultrasonic.[3] The mechanical sensor was an arm hanging from the ceiling with shaft encoders at its joints; Sutherland called it "rather heavy and uncomfortable to use" and wrote that it had been built to have a sure method of measuring head position.[3] The ultrasonic sensor was designed and built at the MIT Lincoln Laboratory by Charles Seitz and Stylianos Pezaris.[3]
| Element | Description |
|---|---|
| Signal | Continuous wave ultrasound |
| Transmitters | Three, attached to the head-mounted optical system, at 37, 38.6 and 40.2 kHz |
| Receivers | Four, mounted in a square array in the ceiling |
| Signal paths | Twelve (each receiver has an amplifier and three filters, one per transmitter frequency) |
| Output | Phase shift of each path read by the computer as a separate five-bit number |
| Ambiguity | About 1/3 inch, the wavelength of 40 kHz sound in air |
Sutherland explained that the team chose continuous wave ultrasound over pulses because inexpensive narrow-band transducers were available, and to avoid confusion from pulsed noise such as typewriters produce, which had caused difficulty for the pulsed ultrasonic Lincoln Wand.[3] The choice introduced a wavelength ambiguity: each of the twelve measurements was precise within a wave but did not tell which wave was being measured. The computer counted major changes in phase to follow motions of more than one wavelength, so the ambiguity appeared as a constant "initialization error", the difference between the computer's first guess of each path length and the true one. Sutherland expected the redundancy of twelve measurements to resolve these errors and reported "encouraging results", but wrote that a full report on the sensor was not yet possible.[3]
The ultrasonic head position sensor was also used in Sutherland's preliminary experiments at the MIT Lincoln Laboratory in late 1966 and early 1967. There, he wrote, it "operated well enough to measure head position for a few minutes before cumulative errors were objectionable".[3] The paper set a target head-tracking resolution of 1/100 of an inch and one part in 10,000 of rotation, and an accuracy on the order of one tenth of an inch.[3]
Comparison with time-of-flight tracking
Acoustic trackers are divided into time-of-flight and phase coherent types.[1][4] A time-of-flight tracker measures how long a short ultrasonic pulse takes to travel from source to sensor; examples named by Mazuryk and Gervautz are the Logitech 6DOF Ultrasonic Head Tracker and the Mattel Power Glove.[1]
| Property | Phase coherent | Pulsed time-of-flight |
|---|---|---|
| Signal | Continuous tone[2] | Brief ultrasonic pulse[2] |
| Quantity measured | Phase shift against a reference signal[1] | Flight time of the pulse[1] |
| Distance output | Relative change; absolute value needs a known start and a cycle count[2] | Absolute distance from each measurement[4] |
| Measurement delay | Continuous, without latency[2] | About 1 ms per foot of range for the sound to travel[2] |
| Multipath | Received amplitude and phase vary unpredictably with position[2] | Largely avoided by using the first pulse to arrive[2] |
Welch and Foxlin, writing in 2002, noted that "all known commercial acoustic ranging systems" worked by timing the flight of a brief ultrasonic pulse.[2] Mazuryk and Gervautz wrote in 1996 that, because the method delivers relative data whose error tends to accumulate over time, development of phase coherent trackers "was relinquished".[1]
Limitations
Phase coherent methods rely on slow movement: the target must move less than one wavelength between updates for the system to keep track of it.[4] In its raw form it is not resilient to Doppler effect.
Because each reading updates the previous one rather than measuring absolute position, phase coherent trackers are subject to error accumulation over time.[4] Sutherland reported cumulative errors of this kind in his 1966 and 1967 experiments, where they became objectionable after a few minutes.[3]
It is at risk of multi-path interference, meaning sound can bounce off of walls and interfere.[5] Multipath means the received signal is the sum of the direct-path signal and one or more reflected signals with longer paths. Walls and objects in a room reflect sound strongly, so the amplitude and phase received from a continuous-wave emitter vary "drastically and unpredictably" as the receiver moves. Welch and Foxlin suggest this may be the reason no successful implementation of the phase coherent approach had been developed. Pulsed systems avoid most of the problem by waiting for the first pulse to arrive, which has taken the direct path unless it is blocked; this works for sound because it travels slowly enough to leave a clear gap between the direct pulse and the first reflection.[2]
Like other ultrasonic trackers, phase coherent systems depend on the speed of sound, which changes with temperature, humidity and air currents. Welch and Foxlin give a rule of thumb of about 0.1 percent change per degree Fahrenheit, or roughly one millimeter of error per degree Fahrenheit at a range of one meter.[2]
Later phase-based acoustic tracking
Phase measurement has reappeared in research on acoustic tracking with smartphones and other consumer devices. In a 2019 CHI paper, Anran Wang and Shyamnath Gollakota of the University of Washington presented MilliSonic, which extracts distance from the phase of frequency modulated continuous wave (FMCW) chirps rather than from a single tone. Its authors state that acoustic multipath had limited earlier 1D acoustic tracking to an accuracy of 5 to 10 mm, which they write "may cause motion sickness with prolonged use" in VR and AR applications. MilliSonic emits 17.5 to 23.5 kHz chirps from a smartphone speaker to a four-microphone array, filters out most reflected paths before reading the phase, and resolves the whole-cycle phase ambiguity (each cycle is about 2 cm of distance) from the distance and speed estimated at the end of the previous chirp. In the authors' tests it reached a median 1D accuracy of 0.7 mm up to 1 m and a median 3D accuracy of 2.6 mm, and the authors present two applications: 3D tracking of VR headsets using a smartphone as the beacon, and 3D tracking of a Google Cardboard viewer with a small microphone array.[6]
Continuous-tone phase tracking is also used for device-free gesture tracking. A survey of acoustic sensing on commodity devices by Chao Cai, Rong Zheng and Jun Luo describes how the phase of a pure tone can be extracted with a coherent receiver and how phase changes give finer timing than cross-correlation. It cites LLAP, a 2016 system that transmits several tones from a mobile device, extracts the phase of echoes from a finger and tracks its displacement through phase shifts, reporting 3.5 mm accuracy for 1D hand movement with under 15 ms latency.[7]
See also
References
- ↑ 1.00 1.01 1.02 1.03 1.04 1.05 1.06 1.07 1.08 1.09 Mazuryk, Tomasz; Gervautz, Michael (1996-02). "Virtual Reality - History, Applications, Technology and Future". Technical report TR-186-2-96-06, Institute of Computer Graphics and Algorithms, Vienna University of Technology. https://www.cg.tuwien.ac.at/research/publications/1996/mazuryk-1996-VRH/mazuryk-1996-VRH-paper.pdf. Retrieved 2026-09-27.
- ↑ 2.00 2.01 2.02 2.03 2.04 2.05 2.06 2.07 2.08 2.09 2.10 2.11 2.12 2.13 Greg Welch, Eric Foxlin (2002). "Motion Tracking: No Silver Bullet, but a Respectable Arsenal". IEEE Computer Graphics and Applications, November/December 2002, pp. 24-38. doi:10.1109/MCG.2002.1046626. https://cise.ufl.edu/research/lok/teaching/ve-s09/papers/cga02_welch_tracking.pdf. Retrieved 2026-09-27.
- ↑ 3.00 3.01 3.02 3.03 3.04 3.05 3.06 3.07 3.08 3.09 3.10 Ivan E. Sutherland (1968-12). "A head-mounted three dimensional display". Proceedings of the AFIPS 1968 Fall Joint Computer Conference, part I, pp. 757-764. doi:10.1145/1476589.1476686. https://www.cise.ufl.edu/research/lok/teaching/ve-s09/papers/sutherland-headmount.pdf. Retrieved 2026-09-27.
- ↑ 4.0 4.1 4.2 4.3 4.4 Gregory Baratoff, Scott Blanksteen (1993). "Tracking Devices". The Encyclopedia of Virtual Environments, Human Interface Technology Laboratory, University of Washington. https://www.hitl.washington.edu/projects/knowledge_base/virtual-worlds/EVE/I.D.1.b.TrackingDevices.html. Retrieved 2026-09-27.
- ↑ 5.0 5.1 Bruce Thomas, Mark Billinghurst (2016-08-23). "COMP 4010 Lecture5 VR Audio and Tracking". SlideShare. University of South Australia. https://www.slideshare.net/slideshow/comp-4010-lecture5-vr-audio-and-tracking/65257828. Retrieved 2026-09-27.
- ↑ Anran Wang, Shyamnath Gollakota (2019). "MilliSonic: Pushing the Limits of Acoustic Motion Tracking". CHI Conference on Human Factors in Computing Systems (CHI 2019), Glasgow. doi:10.1145/3290605.3300248. https://homes.cs.washington.edu/~gshyam/Papers/millisonic.pdf.
- ↑ Chao Cai, Rong Zheng, Jun Luo (2021-08-12). "Ubiquitous Acoustic Sensing on Commodity IoT Devices: A Survey". arXiv:1901.03450 (v2). https://arxiv.org/abs/1901.03450. Retrieved 2026-09-27.