An objective quality assessment method for bit-reduction coding of wideband speech

Area of Science:

• Signal Processing
• Audio Engineering
• Speech Communication

Background:

• Objective speech quality assessment is crucial for optimizing audio codecs.
• Existing methods may not fully capture perceptual aspects of quantization noise in wideband speech.
• Understanding the masking effect of quantization noise is key to accurate quality evaluation.

Purpose of the Study:

• To develop a novel objective quality assessment method for wideband speech coding.
• To incorporate the perceptual masking effect of quantization noise into the quality model.
• To optimize wideband coding system parameters using the proposed objective method.

Main Methods:

• Analysis of speech quality assessment reliability using paired-comparison tests.
• Perceptual study of quantization noise using synthesized speech-noise signals.
• Development of an objective estimation method via multiple regression analysis.
• Inclusion of segmental signal-to-noise ratio, spectral distance, and temporal envelope similarity as factors.

Main Results:

• Quantization noise detection is significantly influenced by masking from the source speech signal.
• A new objective quality estimation method was developed, incorporating key perceptual factors.
• The proposed method demonstrated applicability in optimizing wideband coding system parameters.

Conclusions:

• The developed objective quality assessment method effectively accounts for quantization noise masking in wideband speech.
• This method provides a reliable tool for evaluating and optimizing speech coding systems.
• Further research can explore its application in other audio coding scenarios.