Robust Real-Time Arabic Speech Recognition for AAVs in Adverse Acoustic Conditions Using Lightweight CNNs
Bibliographic record
Abstract
This work introduces an efficient and noise-robust automatic speech recognition (ASR) framework for real-time unmanned aerial vehicle (UAV) control using Modern Standard Arabic (MSA). The proposed approach tackles two major obstacles: the limited availability of Arabic speech resources and the harsh acoustic conditions caused by UAV operation, where propeller, wind, and surrounding noises often impair recognition accuracy. To address these challenges, we enhance a convolutional neural network (CNN) with Squeeze-and-Excitation (SE) attention modules, allowing the model to highlight task-relevant speech cues while attenuating noise-related artifacts. Training was carried out on a purpose-built dataset of 8,800 MSA drone command utterances, with experiments conducted under both clean and noise-augmented conditions. Noise augmentation included three representative disturbances (propeller, wind, and wave) at signal-to-noise ratios (SNRs) of 0, 10, and 20 dB. The system was further validated across seven SNR levels ranging from –5 dB to 30 dB and tested with an unseen noise source originating from a Caterpillar C18 generator. Results show that the clean-trained baseline achieved 98.64% accuracy on the clean test set, while noise-augmented training slightly boosted accuracy to 98.78% and markedly improved robustness. Under 0 dB SNR with the unseen generator noise, the proposed method delivered a 50.55% absolute accuracy improvement over the baseline. With an inference latency of only 0.0256 seconds, the system ensures real-time responsiveness, achieving an effective compromise between recognition performance, computational cost, and resilience to noise, thereby demonstrating its potential for reliable UAV command and control in adverse acoustic conditions.
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 | 0.000 | 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 | 0.002 |
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".