Intelligent Microphone With DLA and RAM for Local Audio Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing microphones require significant data storage and communication bandwidth for processing and transmitting audio data, which can compromise privacy and efficiency.
Innovation Solution
Integration of a Deep Learning Accelerator (DLA) and random access memory in microphones to perform local computations of Artificial Neural Networks (ANNs), reducing the need to transmit raw audio data by generating intelligent outputs directly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If audio data is transmitted and processed externally, then processing capability is improved, but data storage requirements and communication bandwidth increase
Solution Approach 1:
The patent extracts the essential processing function by implementing an Artificial Neural Network (ANN) that converts raw audio data into compact intelligent outputs (such as speech-to-text conversions or recognized patterns). This extraction principle reduces the quantity of data that needs to be stored and transmitted, as only the processed results rather than the complete audio streams are handled externally.
Solution Approach 2:
The patent applies preliminary action by performing audio processing and intelligent recognition locally within the microphone device before data transmission. The ANN processes audio data in real-time, generating condensed outputs that represent the essential information content, thereby reducing subsequent data storage and communication requirements.
2Measurement precision
If audio data is transmitted and processed externally, then processing capability is improved, but communication bandwidth requirements increase
Solution Approach 1:
The patent extracts the essential information from audio data through local ANN processing, converting verbose audio streams into compact intelligent outputs. This extraction significantly reduces the communication bandwidth required, as only the processed results (e.g., recognized speech text or identified patterns) rather than the complete audio data need to be transmitted.
Solution Approach 2:
The patent performs preliminary processing of audio data locally within the microphone using an embedded ANN, generating condensed intelligent outputs before transmission. This preliminary action reduces communication bandwidth requirements by eliminating the need to transmit the entire audio stream, transmitting only the essential processed information instead.
3Loss of information
If local processing is implemented in microphone, then privacy is improved, but device complexity increases
Solution Approach 1:
The patent applies universality by implementing an Artificial Neural Network that can perform multiple processing functions within the microphone device, including speech-to-text conversion, pattern recognition, and audio classification. This multi-functional ANN provides comprehensive privacy protection while managing device complexity through a single integrated processing unit rather than multiple separate components.
Solution Approach 2:
The patent introduces an intermediary ANN layer within the microphone that processes audio data locally before any external transmission. This intermediary processing unit protects privacy by keeping raw audio data local while generating only necessary processed outputs for transmission, effectively mediating between the audio input and external systems.
4Loss of energy
If local processing is implemented in microphone, then energy efficiency is improved, but device complexity increases
Solution Approach 1:
The patent extracts only the essential processing functions needed for intelligent audio recognition and implements them locally using an ANN. By extracting and implementing only the necessary processing capabilities rather than comprehensive audio processing, the system improves energy efficiency while managing device complexity at acceptable levels.
Solution Approach 2:
The patent employs a multi-functional ANN that can perform various processing tasks (speech recognition, pattern matching, audio classification) within a single integrated unit. This universality improves energy efficiency by consolidating multiple processing functions into one system rather than requiring separate processing components, thereby reducing overall energy consumption while managing complexity.
Data Source
AI summary
Systems, devices, and methods related to a Deep Learning Accelerator and memory are described. For example, a microphone may be configured to execute instructions with matrix operands and configured with: a transducer to convert sound waves to electrical signals; an analog to digital converter to generate audio data according to the electrical signals; random access memory to store instructions executable by the Deep Learning Accelerator and store matrices of an Artificial Neural Network; and a controller to store the audio data in the random access memory as an input to the Artificial Neural Network. The Deep Learning Accelerator can execute the instructions to generate an output of the Artificial Neural Network, which may be provided as the primary output of the microphone to a computer system, such as a voice-based digital assistant.