Optimal design of minimum mean-square error noise reduction algorithms using the simulated annealing technique

Area of Science:

• Signal Processing
• Acoustics
• Speech Technology

Background:

• Minimum Mean-Square Error Noise Reduction (MMSE-NR) algorithms combined with Time-Recursive Averaging (TRA) for noise estimation are sensitive to parameter selection.
• Existing noise reduction techniques may not optimally adapt to varying noise conditions or speech characteristics.

Purpose of the Study:

• To develop a systematic optimization method for determining the optimal parameters of MMSE-TRA-NR algorithms.
• To introduce a novel noise reduction (NR) algorithm utilizing linear prediction coding (LPC) for speech signal enhancement.
• To objectively and subjectively evaluate the performance of the proposed optimized algorithm against conventional NR methods.

Main Methods:

• An optimization method using a regression model as the objective function and simulated annealing for parameter search.
• Implementation of a new NR algorithm incorporating linear prediction coding (LPC) for speech correlation extraction.
• Conducting objective and subjective listening tests to assess noise reduction effectiveness.
• Statistical analysis of subjective test results using Analysis of Variance (ANOVA) and Tukey's Honestly Significant Difference (HSD) post hoc test.

Main Results:

• The optimization method successfully identified optimal parameters for the MMSE-TRA-NR algorithm.
• The optimized MMSE-TRA-NR algorithm showed improved performance compared to conventional NR algorithms in both objective and subjective evaluations.
• Statistical analysis confirmed the significant improvements achieved by the proposed algorithm.

Conclusions:

• The proposed optimization method provides an effective approach for tuning MMSE-TRA-NR algorithms.
• The novel NR algorithm incorporating LPC offers a promising alternative for speech enhancement.
• The optimized MMSE-TRA-NR algorithm represents a significant advancement in noise reduction technology for speech processing.