Encoder, Decoder and Methods for Signal-Adaptive Switching of the Overlap Ratio in Audio Transform Coding

Simple SummaryContent extracted from patent full text and abstract with AI.

This patent introduces an improved method for audio transform coding that allows the overlap ratio between audio frames to be adaptively switched based on the properties of the audio signal (such as whether it is transient or stationary/tonal). Traditional codecs like MP3 and AAC use a fixed 50% window overlap during transformation, but this invention allows switching between lower overlaps (5-50%) and higher overlaps (60-100%, typically 75%) using modified lapped transforms and optimized window functions. This adaptive switching reduces quality-degrading artifacts and improves the frequency selectivity for stationary signals while maintaining time resolution during transient segments. The method includes both encoder and decoder designs, facilitating perfect audio reconstruction even during transitions between different overlap ratios.

Use CasesContent extracted from patent full text and abstract with AI.

• Audio codecs for music streaming and download (e.g., in MP3, AAC, or MPEG-H 3D Audio) to improve audio quality at lower bitrates.
• Broadcast and streaming of TV, movies, and live performances with better sound quality and lower delay.
• High-fidelity storage or archival of music and audio content, especially for tonal recordings or classical pieces.
• Hearing aids and communication devices where low delay and high quality are critical.
• Real-time audio communication systems (VoIP, teleconferencing) to maintain clarity during both speech and music.
• Professional audio post-production where transparent audio coding is essential.

BenefitsContent extracted from patent full text and abstract with AI.

• Adaptive overlap ratio optimizes frequency resolution for stationary/tonal parts and time resolution for transients, improving overall audio quality.
• Reduces pre-echo and other coding artifacts that degrade transient or percussive sounds.
• Enables efficient compression without increasing algorithmic delay, benefiting both live and recorded content.
• Maintains compatibility and smooth transitions between different types of lapped transforms, supporting both existing and new audio standards.
• Enables perfect reconstruction of audio signals, meaning the original quality can be preserved during encoding and decoding.
• Flexible implementation that can be deployed in software, hardware, or a combination, and applies to a broad range of audio codecs.

Technical Classifications (CPCs)

Inventors & Applicants

Patent Abstract

A decoder for decoding a plurality of spectral-domain audio samples is provided. The decoder comprises a first decoding module (110) for generating a first group and a second group of time-domain intermediate audio samples from the spectral-domain audio samples. Moreover, the decoder comprises an overlap-adder (130) for overlap-adding the first group of time-domain intermediate audio samples with an overlap of more than 5 % and at most 50 % with the second group of time-domain intermediate audio samples. Furthermore, the decoder comprises a second decoding module (120) for generating a third group and a fourth group of time-domain intermediate audio samples from the spectral-domain audio samples. Moreover, the decoder comprises an output interface (140). The overlap-adder (130) is configured to overlap-add at least the third group of time-domain intermediate audio samples with an overlap of more than 60 % and less than 100 % with the fourth group of time-domain intermediate audio samples. Moreover, the overlap-adder (130) is configured to overlap-add at least the second group and the third group of time-domain intermediate audio samples, or to overlap-add at least the fourth group and the first group of time-domain intermediate audio samples.