Hybrid Feature Optimization for Voice Spoof Detection Using CNN-LSTM
Bibliographic record
Abstract
The objective of this work develops an Automatic Speaker Verification (ASV) system to discern genuine from spoof speech samples.The speech sample features are extracted using Mel-frequency Cepstral Coefficients (MFCC), Constant Q Cepstral Coefficients (CQCC), and Spectrogram feature extraction techniques.MFCC, CQCC, and Spectrogram feature extraction are the most common feature extraction techniques in detecting spoofs in voice samples.However, for detecting voice spoofing using these techniques there is a requirement to improve the accuracy.To improve the accuracy a novel hybrid feature extraction technique is proposed.In this present work, the hybrid features are generated by combining relevant features from the three mentioned feature extraction techniques.These extracted features of the speech samples are fed to the new fused Convolution Neural Network (CNN) model and LSTM Neural Network to improve the performance of the overall system.The data set for evaluating the system is split into training and testing samples.New CNN with LSTM model trains training samples.After completing the training phase, the model is evaluated for testing samples.This work aims to extract the features using all three mentioned and also the generated hybrid feature extraction techniques.The performance of the new CNN with the LSTM model is evaluated through a confusion matrix and ROC curve.Comparing one among all feature extraction techniques, the generated hybrid feature extraction technique provides a better test accuracy of 98.48% and a low Equal Error Rate (EER) of 2.2%.In the end, the new CNN-LSTM architecture achieved the lowest EER among all feature extraction techniques thanks to the hybrid feature extraction approach.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".