Bandwidth Extension Noise Floor Adaptation for Speech Artifacts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing bandwidth extension methods, such as spectral band replication, generate artifacts in high-frequency components due to short-time energy fluctuations, particularly in speech signals, leading to perceptually unacceptable results, especially when bit rates are limited.
Innovation Solution
Adapting the noise floor based on energy distribution data, such as the sibilance parameter derived from LPC coefficients, to adjust the noise floor for voiced speech and sibilant signals, allowing for increased or decreased noise levels to improve perceptual quality, either by transmitting the sibilance parameter or modifying it on the encoder or decoder side.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Volume of moving object
If spectral band replication is used to generate high frequency signals, then bandwidth extension is achieved, but artifacts are generated in high band frequency components due to short-time energy fluctuations
Solution Approach 1:
The patent modifies the noise floor parameter dynamically based on the detected transient condition. When a transient is detected in the base band, the noise floor for the high band is reduced or suppressed. This parameter change prevents the replication of transient artifacts in the extended bandwidth while maintaining the bandwidth extension benefit.
2Object-generated harmful factors
If noise floor is increased to cover sibilant energy fluctuations, then artifacts from sibilants are reduced, but perceptual quality may deteriorate
Solution Approach 1:
The patent applies different noise floor adjustments to different frequency regions and time portions based on local signal characteristics. Sibilant regions receive increased noise floor to mask artifacts, while other regions maintain their original noise characteristics. This localized approach preserves overall perceptual quality while addressing sibilant-specific issues.
Solution Approach 2:
The noise floor is made dynamic rather than static, adapting in real-time to the signal characteristics. The system continuously monitors for sibilant energy fluctuations and adjusts the noise floor accordingly, increasing it only when and where sibilant artifacts are detected, rather than applying a uniform increase across all frequencies and time.
3Difficulty of detecting and measuring
If conventional transient detection is used, then transients are detected, but disturbing artifacts are generated when transients are transformed into higher frequency band
Solution Approach 1:
The patent detects transients (which would normally cause harmful artifacts) and uses this detection information beneficially to suppress or reduce the noise floor in the high band during transient periods. The harmful transient energy in the base band is converted into useful control information that prevents artifact generation in the extended band.
Data Source
Figure 1
Figure 2A~2B
Figure 2C
AI summary
An apparatus (100) for generating bandwidth extension output data (102) for an audio signal (105) comprises a noise floor measurer (110), a signal energy characterizer (120) and a processor (130). The audio signal (105) comprises components in a first frequency band (105a) and components in a second frequency band (105b), the bandwidth extension output data (102) are adapted to control a synthesis of the components in the second frequency band (105b). The noise floor measurer (110) measures noise floor data (115) of the second frequency band (105b) for a time portion (T) of the audio signal (105). The signal energy characterizer (120) derives energy distribution data (125), the energy distribution data (125) characterizing an energy distribution in a spectrum of the time portion (T) of the audio signal (105). The processor (130) combines the noise floor data (115) and the energy distribution data (125) to obtain the bandwidth extension output data (102).