Improvement of Source Number Estimation Method for Single Channel Signal

Area of Science:

• Signal Processing
• Array Signal Processing
• Statistical Signal Processing

Background:

• Accurate source number estimation is crucial for single-channel signal analysis.
• Existing methods like Minimum Description Length (MDL) and Gerschgorin's Disk Estimation (GDE) have limitations, particularly at low signal-to-noise ratios (SNR) and with colored noise.
• MDL excels at low SNR but struggles with colored noise, while GDE handles colored noise but performs poorly at low SNR.

Purpose of the Study:

• To improve the performance of existing source number estimation methods for single-channel signals.
• To address the limitations of MDL and GDE in handling low SNR and colored noise scenarios.
• To propose enhanced algorithms that combine the strengths of both MDL and GDE.

Main Methods:

• Single-channel data is converted to multi-channel data using a delay process.
• Gerschgorin's Disk Estimation (GDE) and Minimum Description Length (MDL) algorithms are applied for source number estimation.
• Diagonal loading technique is employed to enhance the MDL method.
• Jackknife technique is used to optimize the data covariance matrix for the GDE method.

Main Results:

• The enhanced MDL method with diagonal loading shows improved performance.
• The optimized GDE method using the jackknife technique demonstrates better results.
• Simulation results confirm significant performance improvements for both original methods after enhancement.

Conclusions:

• The proposed diagonal loading for MDL and jackknife for GDE significantly enhance source number estimation accuracy.
• These improved methods effectively address the limitations of previous approaches in challenging noise conditions and low SNR environments.
• The study provides more robust and reliable source number estimation techniques for single-channel signal processing.