Evaluation Metrics for a Wideband Signal Processing Framework

Lulu Liu published evaluation metrics on figshare in April 2026 for a proposed wideband signal processing framework. The framework uses a RepViT backbone network to perform detection, recognition, and key parameter extraction on signal spectrograms. Experimental results on a synthetic dataset report a maximum signal recognition rate of 82.43% and normalized root mean squared error for time and frequency parameters.

Use Cases

• Benchmarking new signal detection models based on reported recognition rates (mAcc, aAcc).
• Evaluating parameter extraction accuracy for time and frequency features based on reported NRMSE values.
• Comparing semantic segmentation approaches for signal spectrogram analysis as described in the framework.

Strengths

• Provides concrete performance metrics, including a maximum signal recognition rate (mAcc) of 82.43%.
• Reports specific error ranges for parameter extraction, such as NRMSE for time parameters within 0.3%−2.8%.

Limitations

• The dataset is very small at 5.5 KB, indicating it likely contains only summary metrics, not the underlying raw or processed data.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

Lulu Liu via figshare.

Collection Method

Likely generated from experiments on a large-scale synthetic dataset as described in the associated paper.

Time Range

null

Freshness

Last updated 2026-04-27 17:39:26; freshness should be verified.

Geography

null