Decoder and Encoder for Energy in Bandwidth Extension

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

This patent describes an advanced encoder and decoder system for audio signals, specifically focusing on bandwidth extension. The invention improves how high-frequency (extended-band) audio information is encoded and decoded by predicting the energy of the high-frequency band using linear prediction coefficients and applying an offset that varies depending on the bit-rate. This approach enables efficient and adaptive coding of audio energy for improved sound quality, even in low-bitrate scenarios, and leverages machine learning (DNN) on the decoder to further refine reconstruction quality if needed.

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

• Mobile audio streaming and telecommunication, where high-quality audio is needed at varying and often limited bit-rates (e.g., VoIP, audio calls).
• Music and speech codecs for online media (e.g., music streaming platforms, video calls, podcasts) to enhance audio clarity, especially in narrowband conditions.
• Hearing aids and assistive listening devices that need to reconstruct high-quality audio from band-limited sources.
• Embedded audio applications (e.g., smart speakers, toy electronics, IoT devices) requiring efficient audio processing suitable for hardware with computational or bandwidth constraints.
• Audio restoration or enhancement tools that upgrade historic or bandwidth-constrained recordings.

BenefitsContent extracted from patent full text and abstract with AI.

• Improved audio quality, especially in the high frequencies, even at low bit-rates, leading to better user experiences in speech and music communication.
• Efficient bit-rate usage by predicting and coding only the energy residue, reducing the need for transmitting redundant information.
• Flexibility to operate across a range of bit-rates and coding conditions with a single scheme due to the adaptive offset mechanism.
• Enables use of machine learning on the decoder for further energy prediction refinement without needing additional transmitted data, facilitating advanced enhancements for capable devices.
• Backward-compatible with existing low- and high-band audio encoding techniques, suitable for both speech and general audio coding.
• Reduces computational and storage needs by allowing shared codebooks across different bitrate settings.
• Facilitates easy software or hardware implementation, as methods can be realized in digital signal processors or using computer programs.

Technical Classifications (CPCs)

Inventors & Applicants

Patent Abstract

Encoder (10) for coding a signal comprising a band-limited signal and an extended-band signal, the encoder (10) comprising: a calculator (12) configured to perform energy prediction of the extended-band signal based on LPC coefficients; and a coder (14) configured to encode a residual of the signal using the energy prediction and an offset (o); wherein the offset (o) is dependent on a bit-rate.