Embedded Neural Codec for Low-Latency Hearable Speech Enhancement

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing hearable devices, such as hearing aids and earphones, rely heavily on Bluetooth connectivity and smartphone resources for enhanced capabilities, but this approach limits the scope of these features due to latency and computational burden, particularly in real-time speech enhancement and automatic speech recognition.

Innovation Solution

An embedded neural codec is integrated within hearable devices, combining speech enhancement and automatic speech recognition modules to encode and decode auditory signals efficiently, reducing latency and offloading processing to smartphones for enhanced functionality.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If hearable devices rely on Bluetooth connectivity and smartphone resources for enhanced capabilities, then device complexity is reduced, but latency increases and real-time processing capability deteriorates

Engineering Contradiction:

Improvedevice complexityVSAvoidlatency

Core Design Contradiction:

Device complexityVSLoss of time

Solution Approach 1:

The system segments processing tasks by implementing a hybrid architecture where the hearable device performs critical real-time speech enhancement locally, while transcription and complex analysis are offloaded to the smartphone. This segmentation resolves the contradiction by maintaining low latency for time-sensitive audio processing while reducing overall device complexity through cloud/off-device assistance.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces an intermediary processing architecture where the hearable device's neural codec acts as a mediator between the microphone and the smartphone. The neural codec performs preliminary speech enhancement and compression locally, then transmits processed data to the smartphone for further analysis, thereby reducing latency while sharing computational burden.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Loss of time

If hearable devices perform real-time speech enhancement and transcription locally, then latency is reduced, but computational burden and energy consumption increase

Engineering Contradiction:

ImprovelatencyVSAvoidcomputational burden

Core Design Contradiction:

Loss of timeVSUse of energy by moving object

Solution Approach 1:

The system applies partial action by implementing speech enhancement and compression locally on the hearable device, while deferring transcription and complex NLP tasks to the smartphone. This partial local processing reduces latency for critical audio path while distributing computational burden to avoid excessive energy consumption on the hearable device.

Inventive Principle:

Principle #16Partial or excessive action

Solution Approach 2:

The computational tasks are segmented into two parts: time-critical speech enhancement and compression handled locally by the hearable device's neural codec, and non-time-critical transcription handled by the smartphone. This segmentation optimizes the trade-off between latency and computational energy consumption.

Inventive Principle:

Principle #1Segmentation

3Productivity

If an embedded neural codec is integrated within hearable devices, then real-time processing capability is improved, but device complexity increases

Engineering Contradiction:

Improvereal-time processing capabilityVSAvoiddevice complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The neural codec is designed with multi-functionality, serving both as a compression encoder and a speech enhancement decoder within the hearable device. This universal approach improves real-time processing capability while minimizing the increase in device complexity by consolidating multiple functions into a single integrated component.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent merges the neural codec functionality directly into the hearable device's signal processing pipeline, combining compression and enhancement operations in an integrated architecture. This merging improves real-time processing capability while managing device complexity through unified hardware/software implementation.

Inventive Principle:

Principle #5Merging (Combining)

Data Source

PatentUS20260045265A1Embedded neural codec for low latency speech enhancement and automatic speech recognition in hearables

Publication Date: 2026.02.12 CISCO TECHNOLOGY INC

AI summary

In one embodiment, an example method for using an embedded neural codec for low latency speech enhancement and automatic speech recognition in hearables includes receiving, at a hearable device, an auditory signal and encoding, at the hearable device, the auditory signal into a compressed vector representation of the auditory signal. The method further includes decoding, at the hearable device using a speech enhancement decoder, the compressed vector representation of the auditory signal into denoised speech outputting, from the hearable device, the denoised speech, and transmitting, from the hearable device, an unprocessed vector representation of the auditory signal to a processing device to cause the processing device to decode, using a speech recognition decoder, the compressed vector representation of the auditory signal.