SK
Shivanshu Kumar AnandResearch-Driven AI Architect
LINKEDIN4 PapersCONNECT
← Return Home/Peer-Reviewed & Applied Research
📚 4 Research Publications•Brain-Signal Analysis & Explainable AI

Research Publications & Scientific Benchmarks

Demonstrating deep technical capabilities in neural signal decoding, time-frequency analysis, quantum-inspired feature fusion, and Retrieval-Augmented Generation (RAG) for clinically interpretable AI.

4
Total Papers
99.3%
Peak Accuracy
89.8%
Expert Agreement
🛡️
Academic Integrity & Verification Notice

All research metrics (e.g. 99.31% EEGMAT accuracy and 88.10% Hi3 drowsiness detection) reflect documented peer-reviewed evaluations from the provided manuscripts. IEEE Xplore records (#11280701 & #11295574) are provided with direct source links while publisher verification screens synchronize. No unverified claims or simulated production deployments are made.

INTERACTIVE RESEARCH COMPASS•Where I Am & Where I Am Proceeding

Scientific Trajectory & Technical Evolution

Select Stage:
STAGE 01Published📅 2023 - 2024

Neural & Bio-Morphological Pattern Extraction

⚡ Production AI Translation
🎯 Scientific Problem Tackled

Overcome extreme signal non-stationarity, muscular artifacts in EEG data, and severe diagnostic class imbalances in microscopic medical imaging.

💡 Methodological Breakthrough

Engineered Continuous Wavelet Transform (CWT) multi-scale representations for 2D-CNN diagnostic classification (94.2% ASD detection) and Conditional GANs synthesizing rare leukemia morphologies with zero mode collapse.

📄 Associated Publications & Evidence
EEG-Based Classical Framework for Autism Spectrum Disorder (CNN-ASD)
94.2% Diagnostic Accuracy (IEEE #11280701)
→
Classification of Hematological Malignancies Using Peripheral Blood Smears (GANs)
1:20 Class Imbalance Balanced (IEEE #11295574)
→
🏗️ Production Engineering Translation:

Established our foundational signal pre-processing and artifact removal pipelines now utilized in enterprise audio/telemetry processing.

🔭 Where I Am Proceeding Next:

Cross-subject calibration without needing per-patient baseline recordings.

LIVE INTERACTIVE INFERENCE BENCHMARK

Simulate the Research Architecture in Action

Inspect how raw signal inputs pass through feature extractors, yield confident predictions, and generate grounded RAG rationales.

View Full Methodology Paper →
shivanshu-ai-kernel :: neurology_explainable ai
● PIPELINE LIVE
INPUT TELEMETRY
32-Channel EEG Stream (128Hz) + Workload Cognitive Task Protocol
Theta Band (4-8 Hz)
High (Frontal midline) • Elevated mental effort
Alpha Band (8-12 Hz)
Suppressed (Parietal-occipital) • Active cortical desynchronization
Beta Band (13-30 Hz)
Elevated (Frontal channels) • Heightened cognitive arousal
LATENT INFERENCE CLASSIFIER
Acute Cognitive Stress (High Load)
Reported Accuracy:99.31%
EEGMAT & SAM-40 Benchmark
Latent feature vectors passed forward into the neuro-semantic retrieval engine.
POST-INFERENCE RAG EXPLAINER
Grounded Decision Support
“Frontal midline theta power augmentation coupled with parietal alpha suppression indicates heavy working memory load and acute stress response. Temporal attention weights confirm peak activation occurred at seconds 42–48 during the mental arithmetic task phase.”
✓ 89.8% medical expert blind verification
Literature Context Vector Indexed: Klimesch (1999) — EEG alpha and theta oscillations reflect cognitive memory performance. • Gevins et al. (1998) — Monitoring working memory load using EEG time-frequency analysis.
Filter By Topic:

Showing 4 of 4 Publications

EEG Signal ProcessingCognitive Stress DetectionMultimodal Feature Fusion
Manuscript / Pending Verification

A Multimodal EEG Framework for Cognitive Stress Detection with Retrieval-Augmented Explainability

Rajveer Singh Lalawat, Shivanshu Kumar Anand, Nikhil Kushwaha, Praveen Asthana

🏛️ Peer-Reviewed Research Manuscript (Under Publication Process)•📅 2025–2026

A deep multimodal architecture coupling CNN, Bi-LSTM, and attention mechanisms with Sentence-BERT contextual embeddings and Retrieval-Augmented Generation (RAG) to provide grounded clinical explanations for cognitive stress predictions.

Classification Accuracy
99.31%
Evaluated on EEGMAT Benchmark Dataset
ROC AUC Score
99.98%
EEGMAT Cognitive Stress Dataset
EEG Signal ProcessingDrowsiness DetectionQuantum-Inspired Feature Fusion
Manuscript / Pending Verification

A Deep Learning Architecture with Quantum-Inspired Feature Fusion for Robust EEG-Based Drowsiness Detection

Rajveer Singh Lalawat, Shivanshu Kumar Anand, Praveen Asthana

🏛️ Peer-Reviewed Research Manuscript (Under Publication Process)•📅 2025–2026

A tri-branch deep learning network integrating temporal 1D-CNN-LSTM, spectral STFT 2D-CNN, and quantum-inspired amplitude–phase Hilbert projections to reliably detect human cognitive drowsiness.

Overall Accuracy
88.10%
Reported on benchmark EEG Drowsiness Dataset
Macro F1-Score
0.8667
Harmonic mean across alert and drowsy cohorts
EEG Signal ProcessingAutism Spectrum DisorderContinuous Wavelet Transform
Published

EEG-Based Classical Framework for Autism Spectrum Disorder Diagnosis

Shivanshu Kumar Anand, Rajveer Singh Lalawat

🏛️ IEEE Xplore Digital Library / Peer-Reviewed Conference Proceedings•📅 2024–2025

A specialized CNN-ASD architecture leveraging Continuous Wavelet Transform (CWT) time-frequency scalograms to extract deep spatial-temporal features from non-stationary EEG signals for automated Autism Spectrum Disorder screening.

Classification Accuracy
94.20%
Evaluated on Clinical Benchmark EEG Dataset
Preprocessing Speedup
4.5x
Optimized Wavelet Transform Pipeline
Generative Adversarial NetworksDeep LearningComputer Vision
Published

Classification of Hematological Malignancies Using Peripheral Blood Smear Images

Shivanshu Kumar Anand, Rajveer Singh Lalawat

🏛️ IEEE Xplore Digital Library / Peer-Reviewed Conference Proceedings•📅 2024–2025

A deep learning framework coupling Conditional Generative Adversarial Networks (GANs) with convolutional classification models to synthesize microscopic peripheral blood smear images, overcoming acute class imbalance and accurately classifying blood cancers.

Synthetic Quality
97.10%
FID & Morphological Validation Metric
Minority Subtype Boost
+15.2% F1
Gain over models trained without GAN balancing

The Research Progression: From Neural Decoding to Explainable Enterprise Systems

Shivanshu Kumar Anand's research trajectories form a deliberate narrative: beginning with foundational deep learning signal processing on neurological telemetry (EEG), evolving into multimodal feature learning with quantum-inspired amplitude–phase projections, and culminating in retrieval-augmented clinical explainability.

PHASE 1: SIGNAL PROCESSING

Time-Frequency & Spatial Filtering

Extracting micro-rhythm signals across Theta, Alpha, and Beta bands via STFT spectrograms and 1D/2D CNN encoders.

PHASE 2: NON-LINEAR FUSION

Quantum-Inspired Amplitude–Phase

Employing Hilbert space state superposition to model constructive interference between temporal and spectral branches with low compute latency.

PHASE 3: EXPLAINABILITY & RAG

Retrieval-Augmented Rationales

Augmenting latent neural predictions with verified literature retrieval, reaching 89.8% medical expert agreement on explainability reports.