diff --git a/docs/audio-field.md b/docs/audio-field.md index 888bf8f..bd8112a 100644 --- a/docs/audio-field.md +++ b/docs/audio-field.md @@ -154,6 +154,14 @@ timeline anchors. Current synthetic proof detects all 15 emitted chirps, keeps device runs still depend on loopback capture staying live; the local Scarlett loopback has intermittently stopped advancing during short headless sniffs. +This matcher is a proof, not the final hot path. `BuildChirpletEnergyTrace` +still behaves like a dense sliding matched-filter bank. The current research +note at `research/chirplet-sync-decoder/summary.md` points at the better +receiver shape: design the emitted symbol codebook so a mic can dechirp a +candidate event window and classify symbols by FFT or Goertzel bins. The +transmitter is ours; do not preserve arbitrary symbol shapes if they force the +receiver to spend a CPU core proving they exist. + Next, replace the diagnostic bridge with native audio capture workers that append typed blocks into `Mimir.Runtime`, then expose buffer depth, clock state, delay estimates, and stem routing in Aquarium UI. diff --git a/research/chirplet-sync-decoder/bibliography.md b/research/chirplet-sync-decoder/bibliography.md new file mode 100644 index 0000000..ebda4a8 --- /dev/null +++ b/research/chirplet-sync-decoder/bibliography.md @@ -0,0 +1,49 @@ +# Bibliography: Chirplet Sync Decoder + +Mirrored copies live under `research/chirplet-sync-decoder/mirrors/`. + +## Controlled Chirp / Symbol Decoding + +- H. Hannuna et al., "ChirpMu: Chirp Signal-Based Low-Power Wide-Area Networks," IWQoS 2021. + Source: https://hhannuaa.gitlab.io/papers/iwqos_chirpmu_2021.pdf + Mirror: `mirrors/chirpmu-iwqos-2021.pdf` + Relevance: Uses FFT-band energy over chirp-like symbols and explicitly treats STFT/Hough-style chirp decoding as too complex for constrained devices. + +- J. Zhang et al., "ChirpTransformer: Versatile LoRa Encoding for Low-Power Wide-Area IoT," MobiSys 2024. + Source: https://cse.msu.edu/~caozc/papers/mobisys24-ren.pdf + Mirror: `mirrors/chirptransformer-mobisys-2024.pdf` + Relevance: LoRa-style chirp communication reference for dechirp/FFT symbol detection and chirp-symbol manipulation. + +- D. Vasisht et al., "Choir: Decoding Multiple Transmitters Simultaneously," SIGCOMM 2017. + Source: https://users.ece.cmu.edu/~oyagan/Conferences/SIGCOMM2017.pdf + Mirror: `mirrors/choir-sigcomm-2017.pdf` + Relevance: Demonstrates chirp collision/separation ideas in the LoRa family; useful for future multi-emitter or interference-aware timing beacons. + +## Chirplet Transform / Matching Pursuit + +- J. Cui and W. Wong, "The Adaptive Chirplet Transform and Visual Evoked Potentials," 2017. + Source: https://arxiv.org/pdf/1709.08328 + Mirror: `mirrors/mpact-biosignal-adaptive-chirplet-transform-1709.08328.pdf` + Relevance: Shows the dictionary/matching-pursuit family of chirplet analysis. Useful as a warning: it is expressive, but analysis-oriented and heavier than a controlled sync beacon should be. + +- J. Cui, "MPACT: Matching Pursuit Adaptive Chirplet Transform." + Source: https://github.com/jiecui/mpact + Mirrors: `mirrors/mpact-github.html`, `mirrors/mpact-readme.md` + Relevance: Code reference for adaptive chirplet matching-pursuit structure; not a realtime sync decoder architecture. + +- L. Mann and S. Haykin, "The Chirplet Transform: Physical Considerations," overviewed by later sources. + Source: https://en.wikipedia.org/wiki/Chirplet_transform + Mirror: `mirrors/chirplet-transform-wikipedia.html` + Relevance: Defines the canonical chirplet transform as projection onto chirplet atoms. This legitimizes matched filtering, but does not make dense sliding dictionary search the right machine for Mimir. + +## Fast Chirplet / Detection Notes + +- J. Hu et al., "Fast Chirplet Transform," arXiv:1611.08749. + Source: https://arxiv.org/pdf/1611.08749 + Mirror: `mirrors/fast-chirplet-transform-1611.08749.pdf` + Relevance: Fast chirplet-transform framing; useful for avoiding naive dense scans when a more general chirplet transform is needed. + +- "Edinburgh Maximum Chirplet Transform Code Notes." + Source: https://udrc.eng.ed.ac.uk/sites/udrc.eng.ed.ac.uk/files/attachments/Edinburgh%20Maximum%20Chirplet%20Transform%20Code%20Notes.pdf + Mirror: `mirrors/edinburgh-maximum-chirplet-transform-code-notes.pdf` + Relevance: Practical detection notes around precomputed chirplet windows and CFAR-style detection. Useful for detector architecture, but still broader than Mimir's controlled-code sync problem. diff --git a/research/chirplet-sync-decoder/mirrors/chirplet-transform-wikipedia.html b/research/chirplet-sync-decoder/mirrors/chirplet-transform-wikipedia.html new file mode 100644 index 0000000..434aa28 --- /dev/null +++ b/research/chirplet-sync-decoder/mirrors/chirplet-transform-wikipedia.html @@ -0,0 +1,894 @@ + + + + +Chirplet transform - Wikipedia + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
Jump to content +
+
+
+ + + + +
+
+ + + + + +
+
+
+
+
+
+
+
+
+
+
+ +
+
+
+ +
+
+
+
+
+ +

Chirplet transform

+ +
+ + + +
+
+
+
+
+ +
+
+ + + +
+
+
+
+
+
+ + +
+
+
+
+
+
+ +
From Wikipedia, the free encyclopedia
+
+
+ + +
+
+
Comparison of wave, wavelet, chirp, and chirplet[1]
+
Chirplet in a computer-mediated reality environment.
+

In signal processing, the chirplet transform is an inner product of an input signal with a family of analysis primitives called chirplets.[2][3] +

Similar to the wavelet transform, chirplets are usually generated from (or can be expressed as being from) a single mother chirplet (analogous to the so-called mother wavelet of wavelet theory). +

+ +

Definitions

[edit]
+

The term chirplet transform was coined by Steve Mann, as the title of the first published paper on chirplets. The term chirplet itself (apart from chirplet transform) was also used by Steve Mann, Domingo Mihovilovic, and Ronald Bracewell to describe a windowed portion of a chirp function. In Mann's words: +

+

A wavelet is a piece of a wave, and a chirplet, similarly, is a piece of a chirp. More precisely, a chirplet is a windowed portion of a chirp function, where the window provides some time localization property. In terms of time–frequency space, chirplets exist as rotated, sheared, or other structures that move from the traditional parallelism with the time and frequency axes that are typical for waves (Fourier and short-time Fourier transforms) or wavelets.

+

The chirplet transform thus represents a rotated, sheared, or otherwise transformed tiling of the time–frequency plane. Although chirp signals have been known for many years in radar, pulse compression, and the like, the first published reference to the chirplet transform described specific signal representations based on families of functions related to one another by time–varying frequency modulation or frequency varying time modulation, in addition to time and frequency shifting, and scale changes.[2] In that paper,[2] the Gaussian chirplet transform was presented as one such example, together with a successful application to ice fragment detection in radar (improving target detection results over previous approaches). The term chirplet (but not the term chirplet transform) was also proposed for a similar transform, apparently independently, by Mihovilovic and Bracewell later that same year.[3] +

+

Applications

[edit]
+
(a) In image processing, periodicity is often subject to projective geometry (i.e. chirping that arises from projection). (b) In this image, repeating structures like the alternating dark space inside the windows, and light space of the white concrete, chirp (increase in frequency) towards the right. (c) The chirplet transform is able to represent this modulated variation compactly.
+

The first practical application of the chirplet transform was in water-human-computer interaction (WaterHCI) for marine safety, to assist vessels in navigating through ice-infested waters, using marine radar to detect growlers (small iceberg fragments too small to be visible on conventional radar, yet large enough to damage a vessel).[4][5] +

Other applications of the chirplet transform in WaterHCI include the SWIM (Sequential Wave Imprinting Machine).[6][7] +

More recently other practical applications have been developed, including image processing (e.g. where there is periodic structure imaged through projective geometry),[6][8] +as well as to excise chirp-like interference in spread spectrum communications,[9] in EEG processing,[10] and Chirplet Time Domain Reflectometry.[11] +

+

Extensions

[edit]
+

The warblet transform[12][13][14][15][16][17] is a particular example of the chirplet transform introduced by Mann and Haykin in 1992 and now widely used. It provides a signal representation based on cyclically varying frequency modulated signals (warbling signals). +

+

See also

[edit]
+ +

References

[edit]
+
+
    +
  1. ^ From page 2749 of "The Chirplet Transform: Physical Considerations", S. Mann and S. Haykin, IEEE Transactions on Signal Processing, Volume 43, Number 11, November 1995, pp. 2745–2761. +
  2. +
  3. ^ a b c S. Mann and S. Haykin, "The Chirplet transform: A generalization of Gabor's logon transform", Proc. Vision Interface 1991, 205–212 (3–7 June 1991). +
  4. +
  5. ^ a b D. Mihovilovic and R. N. Bracewell, "Adaptive chirplet representation of signals in the time–frequency plane," Electronics Letters 27 (13), 1159–1161 (20 June 1991). +
  6. +
  7. ^ Mann, Steve, and Simon Haykin. "The chirplet transform: A generalization of Gabor’s logon transform." Vision interface. Vol. 91. 1991. +
  8. +
  9. ^ WaterHCI Part 1: Open Water Monitoring +with Realtime Augmented Reality, IEEE SPICES, INTERNATIONAL CONFERENCE ON SIGNAL PROCESSING, INFORMATICS, COMMUNICATION AND ENERGY SYSTEMS 2022 +(IEEE SPICES 2022), 10 - 12 MARCH, 2022, Nalanchira, Trivandrum, Kerala, India, 6 pages +
  10. +
  11. ^ a b Mann, Steve. "Time-Frequency" Perspectives”." Advances in Machine Vision: Strategies and Applications 32 (1992): 99. +
  12. +
  13. ^ Mann, Steve, et al. "Water-Human-Computer-Interface (WaterHCI): Crossing the Borders of Computation, Clothes, Skin, and Surface." +
  14. +
  15. ^ Mann, Steve, and Simon Haykin. "Adaptive." Optical Engineering 31.6 (1992): 1243-1256. +
  16. +
  17. ^ Bultan, A.; Akansu, A.N. (May 1998), "A novel time-frequency exciser in spread spectrum communications for chirp-like interference", Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), vol. 6, pp. 3265–3268, doi:10.1109/ICASSP.1998.679561, ISBN 0-7803-4428-6 +
  18. +
  19. ^ Cui, J.; Wong, W.; Mann, S. (17 February 2005), "Time–frequency analysis of visual evoked potentials using chirplet transform" (PDF), Electronics Letters, vol. 41, no. 4, pp. 217–218, Bibcode:2005ElL....41..217C, doi:10.1049/el:20056712, archived from the original (PDF) on 2011-07-16, retrieved 2010-07-29 +
  20. +
  21. ^ "Example Programs - National Instruments". Archived from the original on 2012-02-14. Retrieved 2007-12-31. +
  22. +
  23. ^ Mann, Steve, and Simon Haykin. "'Chirplets' and'warblets': novel time-frequency methods." Electronics letters 28, no. 2 (1992): 114-116. +
  24. +
  25. ^ +Mann, S., & Haykin, S. (1992, March). Time-frequency perspectives: the chirplet transform. In Acoustics, Speech, and Signal Processing, 1992. ICASSP-92., 1992 IEEE International Conference on (Vol. 3, pp. 417-420). IEEE. +
  26. +
  27. ^ +Angrisani, L., D'Arco, M., Moriello, R. S. L., & Vadursi, M. (2005). On the use of the warblet transform for instantaneous frequency estimation. Instrumentation and Measurement, IEEE Transactions on, 54(4), 1374-1380. +
  28. +
  29. ^ +Angrisani, L., Arco, M. D., Moriello, R. S. L., & Vadursi, M. (2004, August). Warblet transform based method for instantaneous frequency measurement on multicomponent signals. In Frequency Control Symposium and Exposition, 2004. Proceedings of the 2004 IEEE International (pp. 500-508). IEEE. +
  30. +
  31. ^ +Kazemi, S., Ghorbani, A., Amindavar, H., & Morgan, D. R. (2016). Vital-Sign Extraction Using Bootstrap-Based Generalized Warblet Transform in Heart and Respiration Monitoring Radar System. +
  32. +
  33. ^ +Zelinsky, N. R., & Kleimenova, N. G. Chirplet transform as the useful tool for study the time-frequency structure of geomagnetic pulsations. +
  34. +
+
  • Mann, S.; Haykin, S. (21–26 July 1991), "Adaptive chirplet: An adaptive generalized wavelet-like transform", in Haykin, Simon (ed.), Adaptive Signal Processing, vol. 1565, pp. 402–413, doi:10.1117/12.49794, S2CID 9418542 LEM, Logon Expectation Maximization
  • +
  • Mann, S.; Haykin, S. (1992). "Adaptive chirplet transform". Optical Engineering. 31 (6): 1243–1256. Bibcode:1992OptEn..31.1243M. doi:10.1117/12.57676. introduces Logon Expectation Maximization (LEM) and Radial Basis Functions (RBF) in Time–Frequency space.
  • +
  • Osaka Kyoiku, Gabor, wavelet and chirplet transforms...(PDF)
  • +
  • J. "Richard" Cui, etal, Time–frequency analysis of visual evoked potentials using chirplet transform Archived 2011-07-16 at the Wayback Machine, IEE Electronics Letters, vol. 41, no. 4, pp. 217–218, 2005.
  • +
  • Florian Bossmann, Jianwei Ma, Asymmetric chirplet transform—Part 2: phase, frequency, and chirp rate, Geophysics, 2016, 81 (6), V425-V439.
  • +
  • Florian Bossmann, Jianwei Ma, Asymmetric chirplet transform for sparse representation of seismic data, Geophysics, 2015, 80 (6), WD89-WD100.
  • +
  • L. Angrisani and M. D'Arco, "A measurement method based on a modified version of the chirplet transform for instantaneous frequency estimation," in IEEE Transactions on Instrumentation and Measurement, vol. 51, no. 4, pp. 704-711, Aug. 2002, doi: 10.1109/TIM.2002.803295.
  • +
  • Y. Lu, R. Demirli, G. Cardoso and J. Saniie, "A successive parameter estimation algorithm for chirplet signal decomposition," in IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, vol. 53, no. 11, pp. 2121-2131, November 2006, doi: 10.1109/TUFFC.2006.152.
  • +
  • L. Sorenson, Y. Lu, F. Martinez-Vallina and J. Saniie, "Chirplet Transform Signal Decomposition for Echo Detection and Estimation," 2006 Fortieth Asilomar Conference on Signals, Systems and Computers, Pacific Grove, CA, USA, 2006, pp. 509-512, doi: 10.1109/ACSSC.2006.354800.
  • +
  • Y. Yang, W. Zhang, Z. Peng and G. Meng, "Multicomponent Signal Analysis Based on Polynomial Chirplet Transform," in IEEE Transactions on Industrial Electronics, vol. 60, no. 9, pp. 3948-3956, Sept. 2013, doi: 10.1109/TIE.2012.2206331.
  • +
  • Y. Yang, Z. K. Peng, G. Meng and W. M. Zhang, "Spline-Kernelled Chirplet Transform for the Analysis of Signals With Time-Varying Frequency and Its Application," in IEEE Transactions on Industrial Electronics, vol. 59, no. 3, pp. 1612-1621, March 2012, doi: 10.1109/TIE.2011.2163376.
  • +
  • A. Bhargava and S. Mann, "Adaptive Chirplet Transform-Based Machine Learning for P300 Brainwave Classification," 2020 IEEE-EMBS Conference on Biomedical Engineering and Sciences (IECBES), Langkawi Island, Malaysia, 2021, pp. 62-67, doi: 10.1109/IECBES48179.2021.9398775.
  • +
  • S. Mann, N. Kumar, J. P. Bicalho, M. Sibai and C. Leaver-Preyra, "Adaptive Chirplet Transform-Based Sleep State Detection," 2025 IEEE International Conference on Consumer Electronics (ICCE), Las Vegas, NV, USA, 2025, pp. 1-6, doi: 10.1109/ICCE63647.2025.10929845.
  • +
  • Y. Jiang, W. Chen, M. Li, T. Zhang and Y. You, "Synchroextracting chirplet transform-based epileptic seizures detection using EEG," in Biomedical Signal Processing and Control, vol 68, July 2021, doi: 10.1016/j.bspc.2021.102699
+
[edit]
+ + + + + + + +
+
+ +
+
+ +
+ +
+
+
+
+
+ + + +
+ + +
+
+ +
+
+
+ +
+ + + + \ No newline at end of file diff --git a/research/chirplet-sync-decoder/mirrors/chirpmu-iwqos-2021.pdf b/research/chirplet-sync-decoder/mirrors/chirpmu-iwqos-2021.pdf new file mode 100644 index 0000000..48dba19 Binary files /dev/null and b/research/chirplet-sync-decoder/mirrors/chirpmu-iwqos-2021.pdf differ diff --git a/research/chirplet-sync-decoder/mirrors/chirptransformer-mobisys-2024.pdf b/research/chirplet-sync-decoder/mirrors/chirptransformer-mobisys-2024.pdf new file mode 100644 index 0000000..1a19f72 Binary files /dev/null and b/research/chirplet-sync-decoder/mirrors/chirptransformer-mobisys-2024.pdf differ diff --git a/research/chirplet-sync-decoder/mirrors/choir-sigcomm-2017.pdf b/research/chirplet-sync-decoder/mirrors/choir-sigcomm-2017.pdf new file mode 100644 index 0000000..3556631 Binary files /dev/null and b/research/chirplet-sync-decoder/mirrors/choir-sigcomm-2017.pdf differ diff --git a/research/chirplet-sync-decoder/mirrors/edinburgh-maximum-chirplet-transform-code-notes.pdf b/research/chirplet-sync-decoder/mirrors/edinburgh-maximum-chirplet-transform-code-notes.pdf new file mode 100644 index 0000000..b07339d Binary files /dev/null and b/research/chirplet-sync-decoder/mirrors/edinburgh-maximum-chirplet-transform-code-notes.pdf differ diff --git a/research/chirplet-sync-decoder/mirrors/fast-chirplet-transform-1611.08749.pdf b/research/chirplet-sync-decoder/mirrors/fast-chirplet-transform-1611.08749.pdf new file mode 100644 index 0000000..30ebfce Binary files /dev/null and b/research/chirplet-sync-decoder/mirrors/fast-chirplet-transform-1611.08749.pdf differ diff --git a/research/chirplet-sync-decoder/mirrors/mpact-biosignal-adaptive-chirplet-transform-1709.08328.pdf b/research/chirplet-sync-decoder/mirrors/mpact-biosignal-adaptive-chirplet-transform-1709.08328.pdf new file mode 100644 index 0000000..49c2bf5 Binary files /dev/null and b/research/chirplet-sync-decoder/mirrors/mpact-biosignal-adaptive-chirplet-transform-1709.08328.pdf differ diff --git a/research/chirplet-sync-decoder/mirrors/mpact-github.html b/research/chirplet-sync-decoder/mirrors/mpact-github.html new file mode 100644 index 0000000..a6538cd --- /dev/null +++ b/research/chirplet-sync-decoder/mirrors/mpact-github.html @@ -0,0 +1,1768 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + GitHub - jiecui/mpact: Adaptive Chirplet Transform with Matching Pursuit · GitHub + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
+
+ + +
+ Skip to content + + + + + + + + + + + +
+
+ + + + + + + + + + + + + + + + + + + +
+ +
+ + + + + + + + +
+ + + + + +
+ + + + + + + + + +
+
+
+ + + + + + + + + + +
+ +
+ +
+ +
+ + + + / + + mpact + + + Public +
+ + +
+ +
+ + +
+
+ +
+
+ + + + +
+ + + + + +
+ + + + + + + + + + + + + + + + + + +
+ + + + +

jiecui/mpact

+ + + +

Repository files navigation

Matching Pursuit Based Adaptive Chirplet Transform (MPACT)

+

MPACT is an open-source MATLAB® toolbox that implements multi-component +chirplet signal analysis

+

$$s\left( t\right) = {\left( \frac{\alpha }{\pi }\right)} ^{1/4}\exp \left[ +-\frac{\alpha }{2}{\left( t-t_{0}\right)} ^{2}\right] \times \exp \left[ +j\frac{\beta }{2}{\left( t-t_{0}\right)} ^{2}\right] \exp \left[ j\omega +_{0}\left( t-t_{0}\right) \right]$$

+

using matching pursuit algorithm. The code repository for MPACT is +hosted on GitHub at https://github.com/jiecui/mpact.

+

Installation

+
    +
  1. Copy the files to the directory of your choice, e.g. ~/mpact/
  2. +
  3. Within MATLAB, go the directory where you have copied the files +e.g. >> cd(‘~/mpact/’);
  4. +
  5. Add the directory and its subdirectories into MATLAB search path.
  6. +
  7. Type MATLAB command, i.e. >> doc, to bring up MATLAB help browser.
  8. +
  9. Find “Adaptive Chirplet Transform with Matching Pursuit Toolbox” in +the section of Supplemental Software.
  10. +
  11. Follow the instructions on the screen.
  12. +
+

References

+
    +
  • +

    Cui, J. and D. Wang (2017). "Biosignal Analysis with Matching-Pursuit +Based Adaptive Chirplet." arXive (pre-print +https://arxiv.org/abs/1709.08328)

    +
  • +
  • +

    Cui, J. and W. Wong (2006). "The adaptive chirplet transform and +visual evoked potentials." IEEE Transactions on Biomedical Engineering +53(7): 1378-1384.

    +
  • +
  • +

    Cui, J., Wong, W., & Mann, S. (2005). "Time-frequency analysis of +visual evoked potentials using chirplet transform." Electronics Letters, +41(4), 217-218. (PDF +http://individual.utoronto.ca/jiecui/_private/Time_frequency_chirplet_vep.pdf)

    +
  • +
+

License

+

MPACT is distributed by the GPL v3 Open Source License.

+
+ +
+
+
+
+

About

+ +

+ Adaptive Chirplet Transform with Matching Pursuit +

+ + +

Resources

+ + + +

License

+ + + + + + + + + + + + + + + +

Stars

+ + +

Watchers

+ + +

Forks

+ + + + +
+ +
+
+ + + + + + + +
+
+ + + +

+ Packages +

+ + +
+
+
 
+
+
+
+
 
+
+
+
+
 
+
+ + + + +
+
+
+ + + +
+
+ + +

+ Contributors

+ +
    +
  • +
    +
  • +
  • +
    +
  • +
  • +
    +
  • +
+ + +
+
+
+ + +
+
+

Languages

+
+ + + +
+ + +
+
+ + +
+
+
+ + + + +
+ +
+ +
+
+ +
+ + + + + + + + + + + + + + + + + + + + + + +
+
+
+ + + diff --git a/research/chirplet-sync-decoder/mirrors/mpact-readme.md b/research/chirplet-sync-decoder/mirrors/mpact-readme.md new file mode 100644 index 0000000..c5ada66 --- /dev/null +++ b/research/chirplet-sync-decoder/mirrors/mpact-readme.md @@ -0,0 +1,44 @@ +Matching Pursuit Based Adaptive Chirplet Transform (MPACT) +========================================================== + +**MPACT** is an open-source MATLAB® toolbox that implements multi-component +chirplet signal analysis + +$$s\left( t\right) = {\left( \frac{\alpha }{\pi }\right)} ^{1/4}\exp \left[ +-\frac{\alpha }{2}{\left( t-t_{0}\right)} ^{2}\right] \times \exp \left[ +j\frac{\beta }{2}{\left( t-t_{0}\right)} ^{2}\right] \exp \left[ j\omega +_{0}\left( t-t_{0}\right) \right]$$ + +using matching pursuit algorithm. The code repository for **MPACT** is +hosted on GitHub at https://github.com/jiecui/mpact. + +Installation +------------ + +1. Copy the files to the directory of your choice, e.g. ~/mpact/ +1. Within MATLAB, go the directory where you have copied the files + e.g. >> cd(‘~/mpact/’); +1. Add the directory and its subdirectories into MATLAB search path. +1. Type MATLAB command, i.e. >> doc, to bring up MATLAB help browser. +1. Find “Adaptive Chirplet Transform with Matching Pursuit Toolbox” in + the section of Supplemental Software. +1. Follow the instructions on the screen. + +References +---------- +* __Cui, J.__ and D. Wang (2017). "Biosignal Analysis with Matching-Pursuit + Based Adaptive Chirplet." arXive (pre-print + https://arxiv.org/abs/1709.08328) + +* __Cui, J.__ and W. Wong (2006). "The adaptive chirplet transform and + visual evoked potentials." IEEE Transactions on Biomedical Engineering + 53(7): 1378-1384. + +* __Cui, J.__, Wong, W., & Mann, S. (2005). "Time-frequency analysis of + visual evoked potentials using chirplet transform." Electronics Letters, + 41(4), 217-218. (PDF + http://individual.utoronto.ca/jiecui/_private/Time_frequency_chirplet_vep.pdf) + +License +------- +**MPACT** is distributed by the GPL v3 Open Source License. diff --git a/research/chirplet-sync-decoder/summary.md b/research/chirplet-sync-decoder/summary.md new file mode 100644 index 0000000..f57cc21 --- /dev/null +++ b/research/chirplet-sync-decoder/summary.md @@ -0,0 +1,147 @@ +# Chirplet Sync Decoder: Distilled Research + +## Objective + +Find the right receiver architecture for Mimir's continuous acoustic timing +beacon. The immediate question was whether the current `BuildChirpletEnergyTrace` +and per-symbol matched scoring path is canonical, or whether a standard chirp +decoder trick is missing. + +## Current Mechanism + +The live code currently behaves like a dense sliding matched-filter bank: + +```text +for every audio window: + for every time hop: + for every chirplet symbol: + compute normalized dot product against that symbol kernel + keep peaks + classify symbols around peaks + search code-valid triplets + fit source clock +``` + +This is legal chirplet analysis, but it is not the machine we want. The expensive +chirplet projection is being used as broad discovery. That spends CPU before the +timeline/codebook constraints get a vote. + +## Main Finding + +For a controlled synchronization beacon, the better trick is usually **dechirp +then FFT**, not brute-force symbol scoring. + +Chirp communication systems, especially LoRa-family decoders, exploit the fact +that multiplying a received chirp by the conjugate/reference chirp turns a +frequency-shifted chirp into a near-stationary tone. One FFT then reveals the +symbol as a peak bin. That changes classification from: + +```text +32 full chirplet dot products per event candidate +``` + +to: + +```text +one dechirp multiply + one FFT per event candidate +``` + +If symbols use a shared duration, window, and chirp rate, and encode identity as +frequency offset or bin, the receiver can classify the whole codebook at once. + +## What The Sources Say + +- Chirplet-transform and matching-pursuit literature supports projection onto + chirplet atoms, but that path is general signal analysis. It is expressive and + expensive. +- Fast chirplet-transform work exists because naive chirplet dictionaries are + costly. +- Practical chirp communication systems avoid exhaustive chirp-shape scoring by + designing the waveform so the receiver can dechirp and FFT. +- ChirpMu explicitly contrasts simpler FFT-band chirp symbol decoding with + heavier chirp-parameter methods such as STFT/Hough-style approaches. + +## Design Correction + +The codebook should be redesigned to be receiver-cheap. + +Current shape: + +```text +symbol identity = arbitrary chirp shape: + start band + glide shape + duration + following gap + +receiver = score every symbol kernel +``` + +Better shape: + +```text +symbol identity = FFT-decodable chirp bin: + common chirp duration + common chirp rate + common window + symbol encoded as frequency offset/bin + +receiver = dechirp + FFT peak +``` + +Rhythm/gap can still carry timeline evidence, but it should not be the only +reason symbol identity survives bad microphones. The acoustic symbol should be +cheap and robust by construction. + +## Intended Decoder Machine + +```mermaid +flowchart TD + A["audio stream"] --> B["rolling normalized PCM"] + B --> C["cheap event/onset proposal"] + C --> D["event window"] + D --> E["dechirp by reference chirp"] + E --> F["FFT / Goertzel bin bank"] + F --> G["symbol candidates"] + G --> H["gap and de Bruijn triplet decoder"] + H --> I["canonical timeline anchors"] + I --> J["source clock fit"] +``` + +For Mimir's realtime path, the FFT stage can probably be replaced by a fixed +Goertzel/sliding DFT bank over the 32 symbol bins if the symbol spacing is fixed +and small. That keeps the same invariant: one common dechirp, many cheap bins. + +## What To Cut + +`BuildChirpletEnergyTrace` should not remain the primary acquisition path. It +uses the most expensive operation as the first broad scan. + +The next coherent implementation cut is: + +1. Add a new FFT-decodable chirp-bin codebook beside the current codebook. +2. Render a deterministic de Bruijn timeline with common chirp slope/window and + per-symbol frequency bin. +3. Add a synthetic self-test where dechirp+FFT decodes every emitted symbol and + recovers the same triplet anchors the current matcher recovers. +4. Only then swap live runtime decoding over. + +## Non-Goals + +- Do not keep optimizing dense sliding dot products and call that the final + transform. SIMD helped, but the structure is still wrong. +- Do not add outlier filters to hide weak symbol identity. +- Do not preserve the arbitrary-shape codebook just because it exists. The + transmitter is ours. Make it serve the receiver. + +## Open Questions + +- Best event proposal: amplitude/onset novelty, broadband high-frequency energy, + one reference dechirp energy trace, or a small slope bank. +- FFT size/bin spacing at 44.1 kHz and 48 kHz so all mics can decode without + resampling artifacts creating bin ambiguity. +- Whether to keep one chirp slope or use a tiny slope bank for robustness. +- How musical the emitted pattern can remain when symbol identity is constrained + to FFT bins. +- Whether Goertzel bins beat full FFT for the expected symbol count and event + cadence. diff --git a/state/evidence.jsonl b/state/evidence.jsonl index 31a5bd5..ae75fa9 100644 --- a/state/evidence.jsonl +++ b/state/evidence.jsonl @@ -51,3 +51,4 @@ {"ts": "2026-05-22T15:52:26+00:00", "type": "implementation", "status": "ok", "note": "Restored the repo-local Python persistence helpers from the Epiphany-style helper family after checking VoidBot and sibling repos. Mimir now has tools/localcast_state.py for status/evidence/branch edits and tools/localcast_prepare_compaction.py for pre-compaction state audits; AGENTS.md points future agents at both helpers."} {"ts": "2026-05-22T17:34:45+01:00", "type": "verification", "status": "ok", "note": "Aquarium WASAPI resampler used int multiplication for source.Length * targetSampleRate; a 60s 48k chirplet batch overflowed and queued one 44.1k output frame. Switching to double-length arithmetic restored 2,646,000 output frames and audible chirplets."} {"ts": "2026-05-22T18:08:51+01:00", "type": "verification", "status": "partial", "note": "Release live chirplet decoder test: Scarlett loopback->Scarlett mic produced repeated runtime reports in ~230-335 ms per 2s analysis window, typically 3-5 matched timeline events. Pair tests also produced reports for Kiyo mic. Kiyo Pro and both USB Camera mics showed chirplet energy but failed to form valid triplet anchors in this pass. Six-audio-source fan-out exposed process-probe buffer continuity/overlap failures, with many candidates rejected as insufficient-window."} +{"ts": "2026-05-22T18:34:16+01:00", "type": "research", "status": "ok", "note": "Archived chirplet decoder research under research/chirplet-sync-decoder. Distilled conclusion: current BuildChirpletEnergyTrace is dense sliding matched filtering; for Mimir's controlled sync beacon, redesign symbol codebook for dechirp+FFT/Goertzel classification so code constraints shape the receiver before expensive scoring."}