Journal: Sensors (Basel, Switzerland)

Article Title: Deception Detection from Five-Channel Wearable EEG on LieWaves: A Reproducible Baseline for Subject-Dependent and Subject-Independent Evaluation

doi: 10.3390/s26031027

Figure Lengend Snippet: Basic dataset diagnostics for LieWaves: ( a ) session-level class distribution (27 truth vs. 27 lie sessions); and ( b ) histogram of raw EEG file sizes (kB) across all 54 sessions, indicating broadly consistent recording duration.

Article Snippet: Electroencephalography (EEG) offers non-invasive, millisecond-level access to large-scale cortical dynamics [ ] and has been used to probe cognitive operations such as memory, attention, and cognitive control [ , ] during truthful responding and concealment [ , , ].

Techniques:

Journal: Sensors (Basel, Switzerland)

Article Title: Deception Detection from Five-Channel Wearable EEG on LieWaves: A Reproducible Baseline for Subject-Dependent and Subject-Independent Evaluation

doi: 10.3390/s26031027

Figure Lengend Snippet: Subject 1, session 1: ( a ) 30 s excerpt of an automatic wavelet-packet artifact reduction (ATAR)-preprocessed recording on the five EEG channels (shaded bars mark stimulus-on intervals); and ( b ) PSD for the same session and channels, showing out-of-band attenuation consistent with the 1–45 Hz pass-band (the additional 50 Hz notch is redundant under this range), as well as between-channel spectral differences (log-scale y -axis).

Article Snippet: Electroencephalography (EEG) offers non-invasive, millisecond-level access to large-scale cortical dynamics [ ] and has been used to probe cognitive operations such as memory, attention, and cognitive control [ , ] during truthful responding and concealment [ , , ].

Techniques:

Journal: Sensors (Basel, Switzerland)

Article Title: Deception Detection from Five-Channel Wearable EEG on LieWaves: A Reproducible Baseline for Subject-Dependent and Subject-Independent Evaluation

doi: 10.3390/s26031027

Figure Lengend Snippet: Subject 1, session 1: ( a ) mean band power over 2.0 s automatic wavelet-packet artifact reduction (ATAR)-preprocessed windows, averaged across the five EEG channels for lie and truth segments; and ( b ) mean PSD across the same channels for the same session, with the 1–45 Hz pass-band highlighted (log-scale y -axis).

Article Snippet: Electroencephalography (EEG) offers non-invasive, millisecond-level access to large-scale cortical dynamics [ ] and has been used to probe cognitive operations such as memory, attention, and cognitive control [ , ] during truthful responding and concealment [ , , ].

Techniques:

Journal: Sensors (Basel, Switzerland)

Article Title: Deception Detection from Five-Channel Wearable EEG on LieWaves: A Reproducible Baseline for Subject-Dependent and Subject-Independent Evaluation

doi: 10.3390/s26031027

Figure Lengend Snippet: Architecture of the Residual Temporal Convolutional Network with Squeeze-and-Excitation and Attention (Res-TCN-SE-Attention) model for deception detection from automatic wavelet-packet artifact reduction (ATAR)-preprocessed EEG. The raw waveform branch (top) receives 2.0 s windows (256 samples × 5 channels at 128 Hz) and passes them through Gaussian noise injection, an initial Conv1D layer, and five depthwise-separable residual blocks with squeeze-and-excitation (SE) modulation and progressive temporal down-sampling, followed by attention pooling and a 192-unit dense layer. In parallel, a DWT branch encodes a 175-dimensional wavelet feature vector with a small multi-layer perceptron, and a FEATS branch encodes a 167-dimensional extended spectral–statistical feature vector. The three embeddings are concatenated and passed through a 160-unit dense layer with dropout and a final sigmoid neuron that outputs the deception probability for each window.

Article Snippet: Electroencephalography (EEG) offers non-invasive, millisecond-level access to large-scale cortical dynamics [ ] and has been used to probe cognitive operations such as memory, attention, and cognitive control [ , ] during truthful responding and concealment [ , , ].

Techniques: Injection, Sampling, Plasmid Preparation

Journal: Sensors (Basel, Switzerland)

Article Title: Deception Detection from Five-Channel Wearable EEG on LieWaves: A Reproducible Baseline for Subject-Dependent and Subject-Independent Evaluation

doi: 10.3390/s26031027

Figure Lengend Snippet: Architecture of the Residual Network with Squeeze-and-Excitation blocks (ResNet-SE) model. The network is a three-stage residual 1D-CNN with squeeze-and-excitation (SE) blocks operating on automatic wavelet-packet artifact reduction (ATAR)-preprocessed 3.0 s EEG windows (384 samples × 5 channels at 128 Hz). Each stage comprises two dilated Conv1D residual blocks (64, 96, and 128 filters with kernel sizes 11, 9, and 7, respectively), with temporal down-sampling after the first block in each stage. The final feature maps are aggregated with global average pooling, followed by dropout, a 128-unit dense layer, and a 1-unit sigmoid output for deception versus truth classification.

Article Snippet: Electroencephalography (EEG) offers non-invasive, millisecond-level access to large-scale cortical dynamics [ ] and has been used to probe cognitive operations such as memory, attention, and cognitive control [ , ] during truthful responding and concealment [ , , ].

Techniques: Sampling, Blocking Assay