Voice Recognition ASIC Chip Local Processing Privacy

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current voice recognition systems face challenges in achieving real-time processing and ensuring user privacy due to reliance on cloud computers, which result in slow response times and privacy concerns from transmitting voice data over the internet.

Innovation Solution

A real-time voice recognition apparatus utilizing an ASIC chip and a smartphone, where feature extraction and acoustic model processing are performed in hardware, and lexical/language model processing is done in software, using a GRU-RNN method and parallel processing with a GPU, without the need for a cloud computer, thereby minimizing delay and ensuring privacy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If cloud computer is used for voice recognition, then recognition accuracy is improved (95%), but response time deteriorates due to Internet communication delay

Engineering Contradiction:

Improvevoice recognition accuracyVSAvoidresponse time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent segments the voice recognition system into two parts: acoustic model processing performed in hardware on the ASIC chip, and lexical/language model processing performed in software on the smartphone. This segmentation allows the time-critical acoustic processing to be accelerated in hardware while keeping the system portable and privacy-preserving.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces an ASIC chip as an intermediary device between the smartphone microphone and the processing units. The ASIC chip receives audio input directly from the smartphone and performs acoustic model processing, acting as a dedicated hardware accelerator that reduces the time burden on the main processor and enables real-time operation.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Power

If cloud computer is used for voice recognition, then processing power is improved, but personal privacy deteriorates due to transmission of audio speeches

Engineering Contradiction:

Improveprocessing powerVSAvoidpersonal privacy risk

Core Design Contradiction:

PowerVSObject-affected harmful factors

Solution Approach 1:

The patent extracts the voice recognition processing capability from the cloud environment and embeds it directly into the smartphone device. By taking out the acoustic model processing and implementing it in hardware on the ASIC chip, the system eliminates the need to transmit audio data to external servers, thereby preserving user privacy while maintaining processing power.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The smartphone equipped with the ASIC chip performs voice recognition processing autonomously without requiring external cloud services. The device serves itself by processing audio input locally through the integrated hardware-accelerated acoustic model and software-based lexical/language model, eliminating privacy risks associated with data transmission.

Inventive Principle:

Principle #25Self-service

3Measurement precision

If LSTM-RNN is implemented using FPGA, then voice recognition accuracy is improved, but device complexity increases due to requirement of high-capacity memory chip outside FPGA

Engineering Contradiction:

Improvevoice recognition accuracyVSAvoidsystem complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent merges the acoustic model processing logic and the necessary memory resources into a single ASIC chip. By combining the RNN/LSTM algorithm implementation with integrated memory structures, the system eliminates the need for separate high-capacity external memory chips, thereby reducing device complexity while maintaining the accuracy benefits of sophisticated neural network processing.

Inventive Principle:

Principle #5Merging (Combining)

4Adaptability or versatility

If voice recognition is performed using CPU and GPU in smartphone, then flexibility is improved, but operating speed deteriorates making real-time operation difficult

Engineering Contradiction:

Improvesoftware flexibilityVSAvoidoperating speed

Core Design Contradiction:

Adaptability or versatilityVSSpeed

Solution Approach 1:

The patent replaces the general-purpose software-based processing (CPU/GPU) with a dedicated hardware implementation for the acoustic model processing. By substituting the mechanical/software system with a hardware-accelerated ASIC chip, the system achieves the speed necessary for real-time operation while preserving software flexibility for the lexical/language model processing on the smartphone.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS11183177B2Real-time voice recognition apparatus equipped with ASIC chip and smartphone

Publication Date: 2021.11.23 POSTECH ACADEMY INDUSTRY FOUNDATION

AI summary

The present invention relates to a real-time voice recognition apparatus equipped with an application-specific integrated circuits (ASIC) chip and a smartphone, capable, by using one smartphone and one ASIC chip and without using a cloud computer, of assuring personal privacy, and, due to a short delay time, enabling real-time conversion of voice input signals into text for output. When one DRAM chip is optionally added to the real-time voice recognition apparatus, the number of neural network layers is increased thereby significantly improving accuracy of conversion of voice input signals into text.