# Entropy-Based Concentration and Instantaneous Frequency of TFDs from Cohen’s, Affine, and Reassigned Classes

^{1}

^{2}

^{3}

^{4}

^{*}

## Abstract

**:**

## 1. Introduction

#### 1.1. Time–Frequency Analysis

#### 1.1.1. Analytic Form of a Signal

#### 1.1.2. Instantaneous Frequency

#### 1.1.3. Group Delay

## 2. Affine TFRs with Respect to Cohen’s and Reassigned TFRs

- $k=2$—Affine Wigner representation (extended covariance along straight line paths);
- $k=1/2$—D-Flandrin representation (extended covariance along square-root-hyperbolic paths);
- $k=0$—Bertrand representation (extended covariance along hyperbolic paths);
- $k=-1$—Unterberger representation;
- $k=\pm 5$—Approximate affine Wigner representations (unsmoothed);
- $k=\pm \infty $—Margenau–Hill representation.

#### 2.1. Kernels

#### 2.2. Energy

#### 2.3. Covariance

#### 2.4. Marginals

#### 2.5. Interference Terms

#### 2.6. Reassignment Method

#### 2.7. Affine Class Distributions

#### 2.7.1. Scalogram

#### 2.7.2. Smoothed Pseudo-Affine Wigner Distribution

#### 2.7.3. Unitary Bertrand Distribution

#### 2.7.4. Unterberger Distribution

#### 2.7.5. D-Flandrin Distribution

#### 2.7.6. Other Affine Class Distributions

## 3. Examples and Simulation Results

#### 3.1. Examples of Synthetic Signals

#### 3.1.1. Examples of Mono-Component Noisy Signals

#### 3.1.2. Example of a Multi-Component Noisy Successive Signal

^{−5}and 136.99 × 10

^{−5}, respectively). From the affine class, AMWSC outperformed other TFRs achieving 76.27 × 10

^{−5}and 113.90 × 10

^{−5}for SNRs of 10 dB and 5 dB, respectively. From the reassigned class, for SNRs of 10 dB and 5 dB, RGSP outperformed other reassigned TFRs, achieving MSEs of 9.39 × 10

^{−5}and 39.73 × 10

^{−5}, respectively. When considering intensive noise scenarios, it is interesting to note that SP offers the most accurate IF estimation for 0 dB (304.51 × 10

^{−5}). For comparison, the worst-performing was AAU from the affine class, resulting in an IF estimation MSE of 1774.30 × 10

^{−5}. Similar conclusions can be drawn for the case of an SNR of −5 dB, where SP’s IF estimation MSE was 1125.70 × 10

^{−5}(negligibly outperformed by RPWV with an MSE of 1082.20 × 10

^{−5}). Again, as for 0 dB, AAU was the worst-performing with 2321.90 × 10

^{−5}. Thus, Cohen’s class resulted in the most accurate IF estimation for intensive noise SNRs for the noisy multi-component signal with successive parabolic and linear FM components, while the reassigned class of TFRs was the best-performing for low-noise environments. In both cases, the affine class was outperformed by the other two classes.

#### 3.1.3. Example of a Multi-Component Noisy Concurrent Signal

#### 3.2. Examples of Real-World Signals

#### 3.3. Computational Cost of Analyzed TFRs

^{®}Core™ i5-7300HQ CPU @ 2.50 GHz, 16 GB of DDR4-2666 (1333 MHz) RAM, and NVIDIA GeForce GTX 1050 Ti GPU with 4 GB GDDR5 of dedicated memory (7.9 GB total shared memory).

## 4. Results Discussion

## 5. Conclusions

## Author Contributions

## Funding

## Institutional Review Board Statement

## Informed Consent Statement

## Data Availability Statement

## Acknowledgments

## Conflicts of Interest

## Abbreviations

TFR | Time–Frequency Representation |

TFD | Time–Frequency Distribution |

TSR | Time–Scale Representation |

TSD | Time–Scale Distribution |

IF | Instantaneous Frequency |

TF | Time–Frequency |

FM | Frequency Modulation |

SNR | Signal-to-Noise Ratio |

MSE | Mean Squared Error |

SP | Spectrogram |

SPAWD | Smoothed Pseudo-Affine Wigner Distribution |

SPWV | Smoothed Pseudo-Wigner–Ville |

WV | Wigner–Ville |

AMWSC | Affine Morlet Wavelet Scalogram |

ASPWV | Affine Smoothed Pseudo-Wigner–Ville |

AUB | Affine Unitary Bertrand |

AUU | Affine Active Unterberger |

ADF | Affine D-Flandrin |

RSP | Reassigned Spectrogram |

RGSP | Reassigned Gabor Spectrogram |

RMSC | Reassigned Morlet Scalogram |

RPWV | Reassigned Pseudo-Wigner–Ville |

RSPWV | Reassigned Smoothed Pseudo-Wigner–Ville |

## References

- Gonçalves, P.; Flandrin, P.; Chassande-Mottin, E. Time-Frequency Methods in Time Series Data Analysis. In Proceedings of the Gravitational Wave Data Analysis Workshop 2, Orsay, France, 13–15 November 1997; Davier, M., Hello, P., Eds.; Editions Frontières: Paris, France, 1997; pp. 35–46. [Google Scholar]
- Gröchenig, K. Foundations of Time-Frequency Analysis; Birkhauser Verlag: Basel, Switzerland, 2013. [Google Scholar] [CrossRef]
- Orini, M.; Laguna, P.; Mainardi, L.T.; Bailón, R. Time-Frequency Analysis of Cardiovascular Signals and Their Dynamic Interactions. In Complexity and Nonlinearity in Cardiovascular Signals; Springer International Publishing: Berlin/Heidelberg, Germany, 2017; pp. 257–287. [Google Scholar] [CrossRef] [Green Version]
- Zeevi, Y.; Coifmana, R. (Eds.) Wavelet Analysis and its Applications. In Signal and Image Representation in Combined Spaces; Academic Press: Cambridge, MA, USA, 1998; Volume 7, p. 585. [Google Scholar] [CrossRef]
- Volpato, A.; Collini, E. Time-frequency methods for coherent spectroscopy. Opt. Express
**2015**, 23, 20040–20050. [Google Scholar] [CrossRef] - Hurley, C.; Mclean, J. Wavelet: Analysis and Methods; Wavelet, ETP: Selangor Darul Ehsan, Malaysia, 2018. [Google Scholar]
- Volpato, A.; Collini, E. Optimization and selection of time-frequency transforms for wave-packet analysis in ultrafast spectroscopy. Opt. Express
**2019**, 27, 2975–2987. [Google Scholar] [CrossRef] [PubMed] - Auger, F.; Chassande-Mottin, E. Quadratic Time-Frequency Analysis I: Cohen’s Class; ISTE: Vallsta, Sweden, 2008; pp. 131–163. [Google Scholar] [CrossRef]
- Boashash, B. Time Frequency Signal Analysis and Processing: A Comprehensive Reference; Academic Press: Amsterdam, The Netherlands; Boston, MA, USA, 2016. [Google Scholar]
- Lerga, J.; Member, V.S.E.; Boashash, B. An Efficient Algorithm for Instantaneous Frequency Estimation of Nonstationary Multicomponent Signals in Low SNR. EURASIP J. Adv. Signal Process.
**2011**, 2011, 725189. [Google Scholar] [CrossRef] [Green Version] - Lerga, J.; Sucic, V. Nonlinear IF Estimation Based on the Pseudo WVD Adapted Using the Improved Sliding Pairwise ICI Rule. IEEE Signal Process. Lett.
**2009**, 16, 953–956. [Google Scholar] [CrossRef] - Lerga, J.; Sucic, V. An instantaneous frequency estimation method based on the improved sliding pair-wise ICI rule. In Proceedings of the 10th International Conference on Information Science, Signal Processing and their Applications (ISSPA 2010), Kuala Lumpur, Malaysia, 10–13 May 2010; pp. 161–164. [Google Scholar] [CrossRef]
- Auger, F.; Flandrin, P.; Gonçalves, P.; Lemoine, O. Time-Frequency Toolbox Tutorial; CNRS: Paris, France; Rice University: Houston, TX, USA, 2005. [Google Scholar]
- Mika, D.; Józwik, J. Advanced Time-Frequency Representation in Voice Signal Analysis. Adv. Sci. Technol. Res. J.
**2018**, 12, 251–259. [Google Scholar] [CrossRef] - Feichtinger, H.G.; Strohmer, T. (Eds.) Gabor Analysis and Algorithms Theory and Applications; Birkhäuser: Boston, MA, USA, 1998. [Google Scholar] [CrossRef]
- Berge, E.; Berge, S.M.; Luef, F. The Affine Wigner Distribution. arXiv
**2019**, arXiv:1912.03112. [Google Scholar] [CrossRef] - Gonçalvés, P.; Ovarlez, J.P.; Baraniuk, R. Quadratic Time-Frequency Analysis III: The Affine Class and Other Covariant Classes. In Time-Frequency Analysis; John Wiley & Sons, Ltd.: Hoboken, NJ, USA, 2008; Chapter 7; pp. 193–226. [Google Scholar] [CrossRef]
- Murray, R.; Papandreou-Suppappola, A.; Boudreaux-Bartels, G. A new class of affine higher order time-frequency representations. In Proceedings of the 1999 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP99 (Cat. No.99CH36258), Phoenix, AZ, USA, 15–19 March 1999; Volume 3, pp. 1613–1616. [Google Scholar] [CrossRef]
- Goncalves, P.; Baraniuk, R. Pseudo affine Wigner distributions. In Proceedings of the 1996 IEEE International Conference on Acoustics, Speech, and Signal Processing Conference Proceedings, Atlanta, GA, USA, 9 May 1996; Volume 3, pp. 1423–1426. [Google Scholar] [CrossRef] [Green Version]
- Goncalves, P.; Baraniuk, R. Pseudo affine Wigner distributions: Definition and kernel formulation. IEEE Trans. Signal Process.
**1998**, 46, 1505–1516. [Google Scholar] [CrossRef] [Green Version] - Sandsten, M. Time-Frequency Analysis of Time-Varying Signals and Non-Stationary Processes; Lund University: Lund, Sweden, 2016. [Google Scholar]
- Boashash, B.; Ouelha, S. Designing high-resolution time–frequency and time–scale distributions for the analysis and classification of non-stationary signals: A tutorial review with a comparison of features performance. Digit. Signal Process.
**2018**, 77, 120–152. [Google Scholar] [CrossRef] - Pikula, S.; Beneš, P. A New Method for Interference Reduction in the Smoothed Pseudo Wigner-Ville Distribution. Int. J. Smart Sens. Intell. Syst.
**2014**, 7, 1–5. [Google Scholar] [CrossRef] [Green Version] - Rioul, O.; Flandrin, P. Time-scale energy distributions: A general class extending wavelet transforms. IEEE Trans. Signal Process.
**1992**, 40, 1746–1757. [Google Scholar] [CrossRef] - Hlawatsch, F.; Kozek, W. The Wigner distribution of a linear signal space. IEEE Trans. Signal Process.
**1993**, 41, 1248–1258. [Google Scholar] [CrossRef] - Stankovic, L. An analysis of some time-frequency and time-scale distributions. Ann. Telecommun.
**1994**, 49, 505–517. [Google Scholar] [CrossRef] - Matz, G.; Hlawatsch, F. Wigner distributions (nearly) everywhere: Time–frequency analysis of signals, systems, random processes, signal spaces, and frames. Signal Process.
**2003**, 83, 1355–1378. [Google Scholar] [CrossRef] - Hlawatsch, F.; Mecklenbräuker, W. (Eds.) The Wigner Distribution; Elsevier: Amsterdam, The Netherlands, 1997; p. 459. [Google Scholar]
- Szmajda, M.; Górecki, K.; Mroczka, J. Gabor Transform, Gabor-Wigner Transform and SPWVD as a time-frequency analysis of power quality. In Proceedings of the 14th International Conference on Harmonics and Quality of Power—ICHQP 2010, Bergamo, Italy, 26–29 September 2010; pp. 1–8. [Google Scholar] [CrossRef]
- Boudreaux-Bartels, G. Mixed Time- Frequency Signal Transformations. In Electrical Engineering Handbook; CRC Press LLC: Boca Raton, FL, USA, 2000; Chapter 12; p. 76. [Google Scholar] [CrossRef]
- Flandrin, P.; Goncalvès, P. Geometry of Affine Time–Frequency Distributions. Appl. Comput. Harmon. Anal.
**1996**, 3, 10–39. [Google Scholar] [CrossRef] [Green Version] - Kodera, K.; De Villedary, C.; Gendrin, R. A new method for the numerical analysis of non-stationary signals. Phys. Earth Planet. Inter.
**1976**, 12, 142–150. [Google Scholar] [CrossRef] - Kodera, K.; Gendrin, R.; Villedary, C. Analysis of time-varying signals with small BT values. IEEE Trans. Acoust. Speech Signal Process.
**1978**, 26, 64–76. [Google Scholar] [CrossRef] - Auger, F.; Flandrin, P. Generalization of the reassignment method to all bilineartime-frequency and time-scale representations. In Proceedings of the ICASSP’94—IEEE International Conference on Acoustics, Speech and Signal Processing, Adelaide, Australia, 19–22 April 1994; Volume 4, pp. IV/317–IV/320. [Google Scholar] [CrossRef]
- Auger, F.; Flandrin, P. Improving the readability of time-frequency and time-scale representations by the reassignment method. IEEE Trans. Signal Process.
**1995**, 43, 1068–1089. [Google Scholar] [CrossRef] [Green Version] - Flandrin, P.; Auger, F.; Chassande-Mottin, E. Time-Frequency reassignment: From principles to algorithms. In Applications in Time-Frequency Signal Processing; Electrical Engineering & Applied Signal Processing Series; Papandreou-Suppappola, A., Ed.; CRC Press: Boca Raton, FL, USA, 2002; Volume 10, pp. 179–204. [Google Scholar] [CrossRef] [Green Version]
- Shafi, I.; Jamil, A.; Shah, S.; Kashif, F. Techniques to Obtain Good Resolution and Concentrated Time-Frequency Distributions: A Review. EURASIP J. Adv. Signal Process.
**2009**, 2009, 43. [Google Scholar] [CrossRef] [Green Version] - Abratkiewicz, K.; Samczyński, P.; Fourer, D. A Comparison of the Recursive and FFT-based Reassignment Methods in Micro-Doppler Analysis. In Proceedings of the 2020 IEEE Radar Conference (RadarConf20), Florence, Italy, 21–25 September 2020; pp. 1–6. [Google Scholar] [CrossRef]
- Fitz, K.R.; Fulop, S.A. A Unified Theory of Time-Frequency Reassignment. arXiv
**2009**, arXiv:0903.3080. [Google Scholar] - Gosme, J.; Richard, C.; Goncalves, P. Adaptive diffusion as a versatile tool for time-frequency and time-scale representations processing: A review. IEEE Trans. Signal Process.
**2005**, 53, 4136–4146. [Google Scholar] [CrossRef] [Green Version] - Hainsworth, S.; Macleod, M. Time Frequency Reassignment: A Review and Analysis; Technical Report; Cambridge University Engineering Department: Cambridge, UK, 2003. [Google Scholar]
- Yang, H.; Sun, M.; Xu, Y. A New Time-frequency Combined Method Based on Improved Complete Ensemble EMD and Affine Smooth Pseudo Wigner-Ville Distribution. UPB Sci. Bull. Ser. D Mech. Eng.
**2021**, 83, 71–86. [Google Scholar] - Bertrand, J.; Bertrand, P. A class of affine Wigner functions with extended covariance properties. J. Math. Phys.
**1992**, 33, 2515–2527. [Google Scholar] [CrossRef] - Bertrand, J.; Bertrand, P. Affine time-frequency distributions. In Time-Frequency Signal Analysis—Methods and Applications; Longman-Cheshire: Melbourne, Australia, 1992; Volume 1992, pp. 118–140. [Google Scholar]
- Flandrin, P.; Rioul, O. Affine smoothing of the Wigner-Ville distribution. In Proceedings of the International Conference on Acoustics, Speech, and Signal Processing, Albuquerque, NM, USA, 3–6 April 1990; Volume 5, pp. 2455–2458. [Google Scholar] [CrossRef]
- Ovarlez, J.; Bertrand, J.; Bertrand, P. Computation of affine time-frequency distributions using the fast Mellin transform. In Proceedings of the ICASSP-92: 1992 IEEE International Conference on Acoustics, Speech, and Signal Processing, San Francisco, CA, USA, 23–25 March 1992; Volume 5, pp. 117–120. [Google Scholar] [CrossRef]
- Shenoy, R.; Parks, T. Affine Wigner distributions. In Proceedings of the ICASSP-92: 1992 IEEE International Conference on Acoustics, Speech, and Signal Processing, San Francisco, CA, USA, 23–25 March 1992; Volume 5, pp. 185–188. [Google Scholar] [CrossRef]
- Flandrin, P.; Goncalves, P. Geometry of affine distributions. In Proceedings of the IEEE-SP International Symposium on Time-Frequency and Time-Scale Analysis, Pittsburgh, PA, USA, 9 October 1994; pp. 80–83. [Google Scholar] [CrossRef]
- Murray, R.; Papandreou-Suppappola, A.; Boudreaux-Bartels, G. New higher order affine time-frequency representations. In Proceedings of the IEEE-SP International Symposium on Time-Frequency and Time-Scale Analysis (Cat. No.98TH8380), Victoria, BC, Canada, 4–6 October 1998; pp. 325–328. [Google Scholar] [CrossRef]
- Iribarren, M.; Martin, C.S.; Saavedra, P. Processing of Non-Stationary Vibrations Using the Affine Wigner Distribution. In Signal Analysis and Prediction; Birkhäuser: Boston, MA, USA, 1998; pp. 89–101. [Google Scholar] [CrossRef]
- Gosme, J.; Richard, C.; Goncalves, P. Diffusion equations for adaptive affine distributions. In Proceedings of the 2004 IEEE International Conference on Acoustics, Speech, and Signal Processing, Montreal, QC, USA, 17–21 May 2004; Volume 2, p. ii-629. [Google Scholar] [CrossRef]
- Xiong, G.; Yang, X.N. Pseudo Affine Spectral Correlation Analysis: Definition and the Kernel Formulation. In Proceedings of the 2006 International Conference on Communications, Circuits and Systems, Guilin, China, 25–28 June 2006; Volume 1, pp. 248–252. [Google Scholar] [CrossRef]
- Gavrovska, A.; Bogdanović, V.; Reljin, I.; Reljin, B. Automatic heart sound detection in pediatric patients without electrocardiogram reference via pseudo-affine Wigner–Ville distribution and Haar wavelet lifting. Comput. Methods Programs Biomed.
**2014**, 113, 515–528. [Google Scholar] [CrossRef] [PubMed] - Chapter 17—Getting Started with a Practical and Efficient Time-Frequency Toolbox TFSAP-7.000. In Time-Frequency Signal Analysis and Processing, 2nd ed.; Boashash, B. (Ed.) Academic Press: Oxford, UK, 2016; pp. 967–988. [Google Scholar] [CrossRef]
- Boashash, B.; Ouelha, S. Efficient software platform TFSAP 7.1 and Matlab package to compute Time–Frequency Distributions and related Time-Scale methods with extraction of signal characteristics. Softwarex
**2018**, 8, 48–52. [Google Scholar] [CrossRef] - Pruša, Z.; Søndergaard, P.; Balazs, P.; Holighaus, N. LTFAT: A Matlab/Octave toolbox for sound processing. In Proceedings of the 10th International Symposium on Computer Music Multidisciplinary Research (CMMR), Marseille, France, 15–18 October 2013; pp. 299–314. [Google Scholar]
- Lerga, J.; Saulig, N.; Stanković, L.; Seršić, D. Rule-Based EEG Classifier Utilizing Local Entropy of Time–Frequency Distributions. Mathematics
**2021**, 9, 451. [Google Scholar] [CrossRef] - Sandsten, M.; Brynolfsson, J.; Reinhold, I. The Matched Window Reassignment. In Proceedings of the 2018 26th European Signal Processing Conference (EUSIPCO), Rome, Italy, 3–7 September 2018; pp. 2340–2344. [Google Scholar] [CrossRef]
- Baraniuk, R.; Flandrin, P.; Janssen, A.; Michel, O. Measuring time-frequency information content using the Renyi entropies. IEEE Trans. Inf. Theory
**2001**, 47, 1391–1409. [Google Scholar] [CrossRef] [Green Version] - Hoffmann, U.; Garcia, G.; Vesin, J.M.; Diserens, K.; Ebrahimi, T. A Boosting Approach to P300 Detection with Application to Brain-Computer Interfaces. In Proceedings of the 2nd International IEEE EMBS Conference on Neural Engineering, Arlington, VA, USA, 16–19 March 2005; pp. 97–100. [Google Scholar] [CrossRef] [Green Version]

**Figure 1.**Example of three elementary Gaussian atoms in time, frequency, and TF domains (spectrogram).

**Figure 2.**Example of three elementary Gaussian atoms in the time, frequency, and TF domains (Wigner-Ville) with interferences.

**Figure 4.**TFRs of a noisy mono-component sinusoidal FM signal with estimated IFs for the SNR of 5 dB.

**Figure 5.**TFRs of a noisy multi-component signal with successive parabolic and linear FM components for 5 dB SNR.

**Figure 6.**TFRs of a noisy multi-component signal with concurrent constant and parabolic FM components for the 5 dB SNR.

**Table 1.**Estimation of IF MSE for noisy mono-component linear FM signal for different SNR levels (averaged over 100 noise realizations), cell colors range from green for the best result to red for the worst result in each column respectively, 1 × 10

^{−5}.

10 dB | 5 dB | 0 dB | −5 dB | |
---|---|---|---|---|

SP | 1.58 | 2.39 | 111.24 | 1329.70 |

SPWV | 1.46 | 2.08 | 98.58 | 1320.50 |

WV | 1.64 | 34.19 | 231.40 | 918.89 |

AMWSC | 4.61 | 9.40 | 326.76 | 2039.20 |

ASPWV | 277.84 | 695.28 | 1789.10 | 2647.70 |

AUB | 31.88 | 77.71 | 706.32 | 2093.80 |

AAU | 75.61 | 280.15 | 1667.80 | 3137.30 |

ADF | 66.84 | 181.90 | 1249.60 | 2381.20 |

RSP | 11.88 | 35.94 | 497.45 | 1811.40 |

RGSP | 13.09 | 37.51 | 683.30 | 2014.00 |

RMSC | 22.85 | 88.66 | 609.31 | 1662.40 |

RPWV | 42.16 | 100.22 | 521.22 | 1288.90 |

RSPWV | 1.57 | 11.84 | 343.78 | 1645.20 |

**Table 2.**TFR Rényi entropy for a noisy mono-component linear FM signal for different SNR levels (averaged over 100 noise realizations), cell colors range from green for the best result to red for the worst result in each column respectively.

10 dB | 5 dB | 0 dB | −5 dB | |
---|---|---|---|---|

SP | 15.225 | 15.475 | 15.959 | 16.480 |

SPWV | 14.905 | 15.293 | 15.898 | 16.437 |

WV | 14.864 | 15.600 | 16.343 | 16.641 |

AMWSC | 13.774 | 14.242 | 15.078 | 15.883 |

ASPWV | 15.022 | 15.551 | 15.933 | 16.134 |

AUB | 14.193 | 15.071 | 15.875 | 16.232 |

AAU | 14.690 | 15.516 | 16.199 | 16.446 |

ADF | 14.301 | 15.191 | 15.919 | 16.187 |

RSP | 12.886 | 13.172 | 13.682 | 14.187 |

RGSP | 12.972 | 13.272 | 13.742 | 14.174 |

RMSC | 12.796 | 13.061 | 13.550 | 14.074 |

RPWV | 13.107 | 14.018 | 14.800 | 15.104 |

RSPWV | 12.240 | 12.854 | 13.684 | 14.349 |

**Table 3.**Estimation of IF MSE for a noisy mono-component sinusoidal FM signal for different SNR levels (averaged over 100 noise realizations), cell colors range from green for the best result to red for the worst result in each column respectively, 1 × 10

^{−4}.

10 dB | 5 dB | 0 dB | −5 dB | |
---|---|---|---|---|

SP | 6.88 | 10.28 | 36.54 | 153.62 |

SPWV | 3.09 | 6.05 | 34.86 | 150.80 |

WV | 187.12 | 153.31 | 135.68 | 150.92 |

AMWSC | 17.83 | 26.08 | 73.08 | 197.99 |

ASPWV | 11.58 | 60.63 | 150.89 | 217.60 |

AUB | 92.30 | 97.53 | 134.96 | 185.98 |

AAU | 99.05 | 127.31 | 187.59 | 245.58 |

ADF | 100.06 | 109.18 | 152.09 | 200.02 |

RSP | 1.79 | 8.54 | 72.52 | 197.80 |

RGSP | 1.06 | 8.68 | 90.08 | 217.16 |

RMSC | 7.93 | 24.19 | 89.60 | 197.35 |

RPWV | 2.74 | 13.09 | 65.93 | 145.99 |

RSPWV | 1.92 | 8.08 | 62.18 | 184.94 |

**Table 4.**TFR Rényi entropy for a noisy mono-component sinusoidal FM signal for different SNR levels (averaged over 100 noise realizations), cell colors range from green for the best result to red for the worst result in each column respectively.

10 dB | 5 dB | 0 dB | −5 dB | |
---|---|---|---|---|

SP | 15.947 | 16.111 | 16.381 | 16.654 |

SPWV | 15.724 | 15.973 | 16.300 | 16.586 |

WV | 16.118 | 16.355 | 16.569 | 16.690 |

AMWSC | 15.170 | 15.402 | 15.788 | 16.227 |

ASPWV | 15.416 | 15.871 | 16.080 | 16.240 |

AUB | 15.482 | 15.765 | 16.070 | 16.340 |

AAU | 15.615 | 15.956 | 16.269 | 16.518 |

ADF | 15.499 | 15.783 | 16.058 | 16.278 |

RSP | 13.579 | 13.720 | 13.998 | 14.301 |

RGSP | 13.368 | 13.587 | 13.923 | 14.249 |

RMSC | 13.522 | 13.689 | 13.991 | 14.207 |

RPWV | 14.127 | 14.578 | 14.957 | 15.152 |

RSPWV | 13.650 | 13.890 | 14.226 | 14.488 |

**Table 5.**Estimation of IF MSE for a noisy multi-component signal with successive parabolic and linear FM components for different SNR levels (averaged over 100 noise realizations), cell colors range from green for the best result to red for the worst result in each column respectively, 1 × 10

^{−5}.

10 dB | 5 dB | 0 dB | −5 dB | |
---|---|---|---|---|

SP | 127.12 | 166.02 | 304.51 | 1125.70 |

SPWV | 67.96 | 136.99 | 325.23 | 1160.50 |

WV | 1046.80 | 1068.60 | 1061.80 | 1206.60 |

AMWSC | 76.27 | 113.90 | 350.10 | 1501.40 |

ASPWV | 148.89 | 480.46 | 1292.90 | 1816.10 |

AUB | 684.84 | 783.60 | 1122.60 | 1546.40 |

AAU | 848.84 | 1098.20 | 1774.30 | 2321.90 |

ADF | 750.37 | 910.03 | 1337.60 | 1707.00 |

RSP | 22.18 | 65.28 | 456.63 | 1509.00 |

RGSP | 9.39 | 39.73 | 445.73 | 1513.40 |

RMSC | 9.84 | 45.78 | 524.60 | 1540.00 |

RPWV | 61.37 | 179.57 | 514.64 | 1082.20 |

RSPWV | 42.95 | 107.03 | 475.40 | 1487.00 |

**Table 6.**TFR Rényi entropy for a noisy multi-component signal with successive parabolic and linear FM components for different SNR levels (averaged over 100 noise realizations), cell colors range from green for the best result to red for the worst result in each column respectively.

10 dB | 5 dB | 0 dB | −5 dB | |
---|---|---|---|---|

SP | 19.616 | 19.839 | 20.226 | 20.519 |

SPWV | 19.379 | 19.696 | 20.148 | 20.449 |

WV | 20.097 | 20.301 | 20.518 | 20.637 |

AMWSC | 18.577 | 18.932 | 19.561 | 20.118 |

ASPWV | 19.184 | 19.742 | 20.108 | 20.267 |

AUB | 19.325 | 19.658 | 20.027 | 20.274 |

AAU | 19.542 | 19.897 | 20.258 | 20.478 |

ADF | 19.358 | 19.692 | 20.027 | 20.231 |

RSP | 16.639 | 16.916 | 17.416 | 17.869 |

RGSP | 16.288 | 16.642 | 17.217 | 17.733 |

RMSC | 16.105 | 16.491 | 17.122 | 17.662 |

RPWV | 17.575 | 18.256 | 18.858 | 19.097 |

RSPWV | 16.837 | 17.244 | 17.753 | 18.032 |

**Table 7.**Estimation of IF MSE for a noisy multi-component signal with concurrent constant and parabolic FM components for different SNR levels (averaged over 100 noise realizations), cell colors range from green for the best result to red for the worst result in each column respectively.

10 dB | 5 dB | 0 dB | −5 dB | |
---|---|---|---|---|

(a) First Component, 1 × 10^{−6} | ||||

SP | 5.26 | 22.41 | 1032.10 | 3914.40 |

SPWV | 7.60 | 38.28 | 1017.90 | 3877.20 |

WV | 4375.40 | 6032.50 | 8521.80 | 9397.40 |

AMWSC | 1.29 | 2.15 | 69.36 | 1446.50 |

ASPWV | 3384.30 | 3910.00 | 4724.80 | 5443.90 |

AUB | 4108.60 | 4593.10 | 5160.00 | 5430.20 |

AAU | 3774.00 | 3900.40 | 4066.40 | 3775.90 |

ADF | 4872.90 | 5352.50 | 5718.60 | 5528.00 |

RSP | 227.92 | 404.64 | 1942.20 | 4245.50 |

RGSP | 168.09 | 487.04 | 2243.00 | 4406.40 |

RMSC | 508.97 | 930.73 | 1785.50 | 4107.40 |

RPWV | 3648.10 | 4975.40 | 6204.10 | 6903.10 |

RSPWV | 45.29 | 193.54 | 1584.20 | 3961.30 |

(b) Second Component, 1 × 10^{−5} | ||||

SP | 5.16 | 8.04 | 136.88 | 483.73 |

SPWV | 5.76 | 8.80 | 133.50 | 467.15 |

WV | 237.34 | 306.31 | 405.21 | 447.99 |

AMWSC | 3.62 | 8.81 | 140.19 | 444.00 |

ASPWV | 400.74 | 450.91 | 557.93 | 635.48 |

AUB | 49.20 | 94.12 | 298.59 | 452.07 |

AAU | 19.29 | 54.31 | 273.11 | 445.41 |

ADF | 137.54 | 216.86 | 408.39 | 508.84 |

RSP | 26.17 | 52.88 | 288.19 | 624.45 |

RGSP | 26.73 | 73.70 | 371.42 | 700.51 |

RMSC | 20.90 | 39.49 | 277.02 | 600.25 |

RPWV | 183.87 | 251.13 | 398.47 | 508.63 |

RSPWV | 5.23 | 21.09 | 222.81 | 544.92 |

**Table 8.**TFR Rényi entropy for a noisy multi-component signal with concurrent linear and parabolic FM components for different SNR levels (averaged over 100 noise realizations), cell colors range from green for the best result to red for the worst result in each column respectively.

10 dB | 5 dB | 0 dB | −5 dB | |
---|---|---|---|---|

SP | 16.033 | 16.168 | 16.392 | 16.643 |

SPWV | 15.794 | 16.051 | 16.345 | 16.598 |

WV | 16.232 | 16.453 | 16.625 | 16.698 |

AMWSC | 14.642 | 14.946 | 15.497 | 16.125 |

ASPWV | 15.738 | 16.030 | 16.182 | 16.271 |

AUB | 15.239 | 15.710 | 16.122 | 16.342 |

AAU | 15.260 | 15.780 | 16.233 | 16.483 |

ADF | 15.287 | 15.737 | 16.099 | 16.277 |

RSP | 13.820 | 13.971 | 14.162 | 14.372 |

RGSP | 13.913 | 14.038 | 14.182 | 14.315 |

RMSC | 13.615 | 13.786 | 14.075 | 14.305 |

RPWV | 14.245 | 14.741 | 15.063 | 15.181 |

RSPWV | 13.308 | 13.774 | 14.223 | 14.506 |

**Table 9.**TFR Rényi entropy of an EEG P300 signal for the Cz electrode (averaged over 466 trials), cell colors range from green for the best result to red for the worst result in each column respectively.

Rényi Entropy | |
---|---|

SP | 13.414 |

SPWV | 13.251 |

WV | 14.366 |

AMWSC | 16.689 |

ASPWV | 16.200 |

AUB | 16.300 |

AAU | 16.302 |

ADF | 16.325 |

RSP | 11.369 |

RGSP | 11.230 |

RMSC | 10.178 |

RPWV | 11.782 |

RSPWV | 11.439 |

**Table 10.**Rényi entropy for the 2020 Zagreb earthquake vertical acceleration seismogram, cell colors range from green for the best result to red for the worst result in each column respectively.

Rényi Entropy | |
---|---|

SP | 18.961 |

SPWV | 18.871 |

WV | 19.831 |

AMWSC | 21.526 |

ASPWV | 20.956 |

AUB | 21.350 |

AAU | 21.417 |

ADF | 21.260 |

RSP | 16.636 |

RGSP | 16.441 |

RMSC | 15.084 |

RPWV | 18.099 |

RSPWV | 16.494 |

**Table 11.**Computation time for analyzed TFRs for tested signals (averaged over 100 realizations, SNR of 5 dB), cell colors range from green for the best result to red for the worst result in each column respectively, in milliseconds (ms).

Noisy Mono Component Linear FM | Noisy Mono Component Sinusoidal FM | Noisy Three Component Parabolic and Linear FM | Noisy Two Component Constant and Parabolic FM | EEG P300 Signal | 2020 Zagreb Earthquake | |
---|---|---|---|---|---|---|

Number of time samples | 128 | 128 | 384 | 128 | 128 | 487 |

SP | 1.6448 | 1.4987 | 7.4916 | 1.4124 | 1.4105 | 12.69 |

SPWV | 23.024 | 22.617 | 225.12 | 22.428 | 22.569 | 369.87 |

WV | 1.4677 | 1.3349 | 7.7009 | 1.2996 | 1.2905 | 13.539 |

AMWSC | 10.219 | 9.7795 | 60.675 | 9.7801 | 9.6643 | 93.012 |

ASPWV | 188.92 | 188.19 | 3122.5 | 188.94 | 189.51 | 4356.7 |

AUB | 284.9 | 283.59 | 2110.8 | 281.91 | 283.34 | 2476.9 |

AAU | 749.92 | 751.21 | 6348.1 | 745.8 | 748.24 | 7999 |

ADF | 250.64 | 250.61 | 1797.7 | 250.28 | 251.04 | 2049.4 |

RSP | 19.526 | 19.205 | 170.08 | 19.449 | 19.203 | 275.86 |

RGSP | 16.126 | 15.649 | 145.11 | 15.565 | 13.163 | 192.5 |

RMSC | 275.13 | 268.87 | 2650.7 | 273.55 | 269.98 | 4207.1 |

RPWV | 3.7773 | 3.4253 | 24.559 | 3.4197 | 3.4948 | 41.555 |

RSPWV | 44.623 | 43.586 | 412.51 | 43.931 | 43.878 | 670.22 |

**Table 12.**Computational time for analyzed TFRs for a noisy linear FM signal with different numbers of time samples (averaged over 100 realizations, SNR of 5 dB), cell colors range from green for the best result to red for the worst result in each column respectively, in milliseconds (ms).

Number of Time Samples | 32 | 64 | 128 | 256 | 512 |
---|---|---|---|---|---|

SP | 0.53308 | 0.68926 | 1.4145 | 3.6583 | 10.057 |

SPWV | 1.9574 | 5.7625 | 22.275 | 92.194 | 407.77 |

WV | 0.38185 | 0.47453 | 1.2714 | 2.968 | 11.174 |

AMWSC | 2.2736 | 3.9472 | 9.6087 | 28.944 | 104.7 |

ASPWV | 26.755 | 47.652 | 179.4 | 455.84 | 4537.6 |

AUB | 46.272 | 103.14 | 278.18 | 791.83 | 2500.5 |

AAU | 99.284 | 257.56 | 727.06 | 2276.1 | 8031.1 |

ADF | 42.68 | 94.99 | 244.54 | 681.51 | 2116.2 |

RSP | 2.2171 | 5.4668 | 18.87 | 70.807 | 285.32 |

RGSP | 2.1111 | 5.0283 | 15.18 | 51.456 | 201.96 |

RMSC | 16.188 | 63.321 | 261.1 | 1110.3 | 4600.3 |

RPWV | 0.99658 | 1.3783 | 3.3517 | 9.9509 | 38.288 |

RSPWV | 3.8662 | 11.514 | 42.46 | 169.57 | 721.94 |

**Table 13.**Computational time for analyzed TFRs of different dimensions for a noisy linear FM signal with 128 time samples (averaged over 100 realizations, SNR of 5 dB), cell colors range from green for the best result to red for the worst result in each column respectively, in milliseconds (ms).

TFR Dimension | $32\times 32$ | $64\times 64$ | $128\times 128$ | $256\times 256$ | $512\times 512$ |
---|---|---|---|---|---|

SP | 1.4193 | 1.3427 | 1.4466 | 1.6453 | 2.0954 |

SPWV | 23.058 | 22.478 | 22.415 | 23.576 | 24.098 |

WV | 1.0921 | 1.0986 | 1.2774 | 1.3618 | 1.5274 |

AMWSC | 3.1886 | 5.0576 | 9.607 | 19.048 | 37.809 |

ASPWV | 34.879 | 57.728 | 101.24 | 188.75 | 447.8 |

AUB | 19.101 | 28.757 | 48.061 | 88.604 | 160.95 |

AAU | 26.891 | 41.811 | 73.08 | 126.18 | 228.62 |

ADF | 16.001 | 25.113 | 43.079 | 75.009 | 141.32 |

RSP | 7.2236 | 11.604 | 19.851 | 38.007 | 72.442 |

RGSP | 5.8219 | 9.1319 | 16.085 | 29.338 | 57.913 |

RMSC | 66.308 | 134.42 | 269.46 | 542.5 | 1129.2 |

RPWV | 2.8339 | 2.7727 | 3.3549 | 4.8437 | 8.2814 |

RSPWV | 30.529 | 34.782 | 42.888 | 59.86 | 95.77 |

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |

© 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Bačnar, D.; Saulig, N.; Petrijevčanin Vuksanović, I.; Lerga, J.
Entropy-Based Concentration and Instantaneous Frequency of TFDs from Cohen’s, Affine, and Reassigned Classes. *Sensors* **2022**, *22*, 3727.
https://doi.org/10.3390/s22103727

**AMA Style**

Bačnar D, Saulig N, Petrijevčanin Vuksanović I, Lerga J.
Entropy-Based Concentration and Instantaneous Frequency of TFDs from Cohen’s, Affine, and Reassigned Classes. *Sensors*. 2022; 22(10):3727.
https://doi.org/10.3390/s22103727

**Chicago/Turabian Style**

Bačnar, David, Nicoletta Saulig, Irena Petrijevčanin Vuksanović, and Jonatan Lerga.
2022. "Entropy-Based Concentration and Instantaneous Frequency of TFDs from Cohen’s, Affine, and Reassigned Classes" *Sensors* 22, no. 10: 3727.
https://doi.org/10.3390/s22103727