Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 53 of the 55 references it cites were found in a registry or an index; 9 of the 13 sources its claims cite were checked for support.

Translation-Invariant Zero-Phase Wavelet Methods for Feature Extraction in Terahertz Time-Domain Spectroscopy

Open access

Abstract

Wavelet transform is an important tool in the computational signal processing of terahertz time-domain spectroscopy (THz-TDS) measurements. Despite its prevalence, the effects of using different forms of wavelet transforms in THz-TDS studies have not been investigated. In this paper, we explore the implications of using the maximal overlap discrete wavelet transform (MODWT) versus the well-known discrete wavelet transform (DWT). We demonstrate that the spectroscopic features extracted using DWT can vary over different overlapping frequency ranges. On the contrary, MODWT is translation-invariant and results in identical features, regardless of the spectral range used for its implementation.We also demonstrate that the details coefficients obtained by the multiresolution analysis (MRA) using MODWT are associated with zero-phase filters. In contrast, DWT details coefficients suffer from misalignments originated from the down- and upsampling operations in DWT pyramid algorithm. Such misalignments have adverse effects when it is critical to retain the exact location of the absorption lines. We study the differences of DWT and MODWT both analytically and experimentally, using reflection THz-TDS measurements of α-lactose monohydrate. This manuscript can guide the researchers to select the right wavelet analysis tool for their specific application of the THz spectroscopy.

Overview

CoChat Summary

TL;DR

This paper analytically and experimentally compares DWT and MODWT for feature extraction in THz-TDS, using reflection measurements of α-lactose monohydrate. It finds MODWT is translation-invariant and yields zero-phase details coefficients, while DWT features depend on spectral range and can be misaligned.

Key findings

  • MODWT wavelet and details coefficients extracted the α-lactose resonance at 0.53 THz regardless of overlapping spectral ranges (0.30–1.10, 0.31–1.11, 0.32–1.12, and 0.33–1.13 THz), whereas DWT did not capture it for the 0.30–1.10 and 0.32–1.12 THz ranges.
  • MODWT details coefficients were identical across different frequency ranges down to four decomposition levels, demonstrating translation invariance; DWT details over f=0.3–1 THz and f=0.4–1.1 THz were not alike.
  • DWT details coefficients showed slight misalignment at the α-lactose resonance at 1.40 THz due to down- and upsampling, while MODWT details aligned perfectly; both DWT and MODWT details are associated with zero-phase filters, but DWT still suffers sampling-related mismatches.

Methods

  • Analytical derivation comparing the DWT pyramid algorithm with downsampling against the MODWT pyramid algorithm without downsampling, including zero-phase multiresolution analysis via cross-correlation.
  • Reflection THz-TDS measurements using a TERA-ASOPS spectrometer with two 1560 nm mode-locked femtosecond lasers detuned by Δf=100 Hz, fpump=100 MHz, effective bandwidth f=0.3–1.6 THz, incidence angle θi=35°, and focal spot size approximately 1.2 mm.
  • Sample preparation: α-lactose monohydrate mixed 1:1 with HDPE, pressed at 3000 PSI for about 30 min into pellets approximately 4 mm thick and 50 mm diameter; average amplitude spectrum of ten disjoint measurements, deconvolved by a reference mirror, detrended, Hann-windowed, zero-padded for J=8 DWT levels, and analyzed with the LA(8)/sym4 wavelet in MATLAB.

Limitations

  • The paper states that MODWT advantages come at a slightly higher computational cost, which it describes as affordable for many applications; no explicit limitations section is included.
  • Evident from the text: experiments use only α-lactose monohydrate and one wavelet filter (LA(8)/sym4), so generalizability to other materials or wavelet filters is not tested.
  • Evident from the text: measurements are reflection THz-TDS over f=0.3–1.6 THz; transmission geometry or other spectral ranges are not examined.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3390/s22062305

Cites

53 of 53 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 (a) The block diagram of the DWT pyramid algorithm. The input signal x(n) is filtered recursively by a set of high-pass wavelet and low-pass scaling filters. The output of each filter is downsampled by a factor of two. (b) The block diagram of the MODWT pyramid algorithm. In MODWT, the sampling rate is the same at all levels of decomposition. However, the (j+1)th-level filters are upsampled by a factor of two with respect to the jth-level filters.
Figure 1 (a) The block diagram of the DWT pyramid algorithm. The input signal x(n) is filtered recursively by a set of high-pass wavelet and low-pass scaling filters. The output of each filter is downsampled by a factor of two. (b) The block diagram of the MODWT pyramid algorithm. In MODWT, the sampling rate is the same at all levels of decomposition. However, the (j+1)th-level filters are upsampled by a factor of two with respect to the jth-level filters.CC BY · PubMed Central
Figure 2 (a) The schematic of the measurement setup. The specular reflections are measured at θi=θr=35^(∘). (b) An example THz pulse reflected from an α-lactose pellet. (c) The spectral amplitude (SA) of α-lactose over the frequency range of f=0.3–1.13 THz. A red arrow marks the zero-crossing at 0.53 THz, which is a characteristic resonance mode of α-lactose. (d) The vertically-offset fourth-level DWT wavelet coefficients of SA, applied in the frequency domain. Each line in (d) is composed of the wavelet coefficients calculated from a different spectral range, i.e., f=0.30–1.10 THz in the purple trace, f=0.31–1.11 THz in the red trace, f=0.32–1.12 THz in the blue trace, and f=0.33–1.13 THz in the green trace. The lower bound of the frequency ranges are emphasized using vertical lines at the beginning of each trace. Each inset box shaded with the same color shows the overlap between the wavelet filter and the resonant signature at 0.53 THz in the corresponding reflection spectra. In each case, it can be noted that whether the resonance at 0.53 THz is resolved in T⁻³W4 depends on the overlap between the spectral feature and the DWT wavelet filter.
Figure 2 (a) The schematic of the measurement setup. The specular reflections are measured at θi=θr=35^(∘). (b) An example THz pulse reflected from an α-lactose pellet. (c) The spectral amplitude (SA) of α-lactose over the frequency range of f=0.3–1.13 THz. A red arrow marks the zero-crossing at 0.53 THz, which is a characteristic resonance mode of α-lactose. (d) The vertically-offset fourth-level DWT wavelet coefficients of SA, applied in the frequency domain. Each line in (d) is composed of the wavelet coefficients calculated from a different spectral range, i.e., f=0.30–1.10 THz in the purple trace, f=0.31–1.11 THz in the red trace, f=0.32–1.12 THz in the blue trace, and f=0.33–1.13 THz in the green trace. The lower bound of the frequency ranges are emphasized using vertical lines at the beginning of each trace. Each inset box shaded with the same color shows the overlap between the wavelet filter and the resonant signature at 0.53 THz in the corresponding reflection spectra. In each case, it can be noted that whether the resonance at 0.53 THz is resolved in T⁻³W4 depends on the overlap between the spectral feature and the DWT wavelet filter.CC BY · PubMed Central
Figure 3 (a) The SA of α-lactose over the frequency range of f=0.3–1.13 THz as the control sample. (b) The vertically-offset fourth-level MODWT wavelet coefficients of the SA signal, applied in the frequency domain. Each line in (b) is calculated over a different frequency range, including f=0.30–1.10 THz in the purple line, f=0.31–1.11 THz in the red line, f=0.32–1.12 THz in the blue line, and f=0.33–1.13 THz in the green line. The lower bound of the frequency ranges are emphasized using vertical lines at the beginning of each trace. Each inset box, shaded with the same color, shows the overlap between the wavelet filter (dashed line) and the resonant signature (solid line) at 0.53 THz in the corresponding reflection spectra. (c) The vertically-offset fourth-level DWT and MODWT wavelet and details coefficients of the SA signal. Panel (d) shows the SA of α-lactose over the entire measurement bandwidth of f=0.3–1.6 THz. In (e) are the vertically-offset third-level DWT and MODWT details vectors of the SA signal in (d). Panel (f) zooms into the details coefficients around 1.4 THz, revealing that the DWT details vector (blue line) is slightly misaligned with respect to the resonant frequency, while MODWT details vector (red line) maintains the exact spectral location. The vertical dashed lines in (a–f) delineate the resonances of α-lactose.
Figure 3 (a) The SA of α-lactose over the frequency range of f=0.3–1.13 THz as the control sample. (b) The vertically-offset fourth-level MODWT wavelet coefficients of the SA signal, applied in the frequency domain. Each line in (b) is calculated over a different frequency range, including f=0.30–1.10 THz in the purple line, f=0.31–1.11 THz in the red line, f=0.32–1.12 THz in the blue line, and f=0.33–1.13 THz in the green line. The lower bound of the frequency ranges are emphasized using vertical lines at the beginning of each trace. Each inset box, shaded with the same color, shows the overlap between the wavelet filter (dashed line) and the resonant signature (solid line) at 0.53 THz in the corresponding reflection spectra. (c) The vertically-offset fourth-level DWT and MODWT wavelet and details coefficients of the SA signal. Panel (d) shows the SA of α-lactose over the entire measurement bandwidth of f=0.3–1.6 THz. In (e) are the vertically-offset third-level DWT and MODWT details vectors of the SA signal in (d). Panel (f) zooms into the details coefficients around 1.4 THz, revealing that the DWT details vector (blue line) is slightly misaligned with respect to the resonant frequency, while MODWT details vector (red line) maintains the exact spectral location. The vertical dashed lines in (a–f) delineate the resonances of α-lactose.CC BY · PubMed Central
Figure 4 The vertically-offset fourth-level DWT details coefficients, (a), are compared to the the same vertically-offset fourth-level MODWT details coefficients in (b). Each trace in (a,b) is calculated over a different frequency range, including f=0.30–1.10 THz in the purple traces, f=0.31–1.11 THz in the red traces, f=0.32–1.12 THz in the blue traces, and f=0.33–1.13 THz in the green traces. In (c,d) the amplitude of all DWT details coefficients are calculated over f=0.3–1 THz, (c), and f=0.4–1.1 THz, (b) for a decomposition level of J=4. In (e,f), the amplitude of all MODWT details coefficients are calculated over f=0.3–1 THz and f=0.4–1.1 THz, respectively, for a decomposition level of J=4. The horizontal white dashed lines separate the levels of decomposition. The details coefficients at each panel are min–max-normalized between zero and one.
Figure 4 The vertically-offset fourth-level DWT details coefficients, (a), are compared to the the same vertically-offset fourth-level MODWT details coefficients in (b). Each trace in (a,b) is calculated over a different frequency range, including f=0.30–1.10 THz in the purple traces, f=0.31–1.11 THz in the red traces, f=0.32–1.12 THz in the blue traces, and f=0.33–1.13 THz in the green traces. In (c,d) the amplitude of all DWT details coefficients are calculated over f=0.3–1 THz, (c), and f=0.4–1.1 THz, (b) for a decomposition level of J=4. In (e,f), the amplitude of all MODWT details coefficients are calculated over f=0.3–1 THz and f=0.4–1.1 THz, respectively, for a decomposition level of J=4. The horizontal white dashed lines separate the levels of decomposition. The details coefficients at each panel are min–max-normalized between zero and one.CC BY · PubMed Central

Verification

Claims the paper makes

5 Valid · 0 Not valid · 1 Not checked · 24 Own results and interpretations

18 more claims

Create a free account to see all 30 claims and every source verdict.

See all claims — free account
  • The quote only partly states this claim. 1
  • It may state two separate claims as one. 6
  • Probably not a substantial claim. 1

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

1 of 1

Citations

Cites

Show all 55 references

Cited by

See all 22

Metadata

Authors
Khani, M. E., & Arbab, M. H.
Affiliations
Stony Brook University (US)
Corresponding author
Arbab, Mohammad Hassan
Venue
Sensors
Publisher
MDPI AG
Published
2022-03-16
Volume
22
Issue
6
Pages
2305
ISSN
1424-8220 (ISSN-L)
OpenAlex
W4221127549
Cited by
22
References
55
Citations by year
2026: 2 · 2025: 4 · 2024: 6 · 2023: 2 · 2022: 8
FWCI
1.37
Type
journal article
Language
en
Open access
Locations
  • Sensors (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
National Institute of General Medical Sciences (GM112693)
Topics
  • Physical Sciences › Engineering › Electrical and Electronic Engineering › Terahertz technology and applications
  • Physical Sciences › Mathematics › Applied Mathematics › Mathematical Analysis and Transform Methods
  • Physical Sciences › Chemistry › Spectroscopy › Spectroscopy and Laser Applications
Keywords
terahertz time-domain spectroscopy, maximal overlap discrete wavelet transform, discrete wavelet transform, translation invariance, zero-phase filters, feature extraction, multiresolution analysis, detail coefficients, reflectance spectroscopy, alpha-lactose monohydrate
MeSH terms
Algorithms, Signal Processing, Computer-Assisted, Terahertz Spectroscopy, Wavelet Analysis

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Translation-Invariant Zero-Phase Wavelet Methods for Feature Extraction in Terahertz Time-Domain Spectroscopy. https://app.cochat.ai/papers/10.3390/s22062305
The paper
Khani, M. E., & Arbab, M. H. (2022). Translation-Invariant Zero-Phase Wavelet Methods for Feature Extraction in Terahertz Time-Domain Spectroscopy. Sensors. https://doi.org/10.3390/s22062305

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.