Cyclic Buffer Queue for Speech Recognition Accuracy

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional smart devices face accuracy issues in speech recognition due to the isolation of speech data between wakeup and recognition engines, leading to partial content interception when users speak wakeup words together with intended commands, resulting in incorrect recognition results.

Innovation Solution

A method involving a client that uses a speech wakeup engine for recognition, stores data in a cyclic buffer queue, determines the interception starting position, and sends buffer data to a speech recognition engine for accurate processing, including secondary recognition if necessary, to ensure complete content is recognized.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If speech data is completely isolated between wakeup engine and recognition engine, then wakeup recognition accuracy is improved, but speech recognition completeness deteriorates

Engineering Contradiction:
Improvewakeup recognition accuracyVSAvoidspeech recognition completeness
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

The patent segments speech data into different blocks and uses a cyclic buffer queue to separate wakeup recognition data from continuous speech recognition data. This allows the wakeup engine to process specific wakeup word blocks while the recognition engine processes other blocks, maintaining both wakeup accuracy and speech completeness.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces a cyclic buffer queue as an intermediary data structure between the wakeup engine and recognition engine. This mediator allows speech data to be stored and selectively retrieved, enabling the system to determine whether each data block should be processed by the wakeup engine or recognition engine based on content analysis.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Speed

If wakeup recognition is performed first, then wakeup detection speed is improved, but speech recognition accuracy deteriorates due to partial content interception

Engineering Contradiction:
Improvewakeup detection speedVSAvoidspeech recognition accuracy
Core Design Contradiction:
SpeedVSMeasurement precision

Solution Approach 1:

The patent performs preliminary wakeup recognition on speech data blocks before full speech recognition. The cyclic buffer queue enables the system to identify wakeup words in advance and mark corresponding data blocks, allowing the recognition engine to skip already-processed wakeup content and focus on remaining speech portions.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent dynamically adjusts the processing path based on wakeup recognition results. When wakeup words are detected in a data block, the system dynamically routes that block to the wakeup engine and excludes it from recognition engine processing, while other blocks continue to the recognition engine, creating a flexible adaptive processing flow.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentUS10923119B2Speech data processing method and apparatus, device and storage medium
Publication Date: 2021.02.16 BAIDU ONLINE NETWORK TECH (BEIJIBG) CO LTD
  • US10923119B2 patent drawing
  • US10923119B2 patent drawing
  • US10923119B2 patent drawing

AI summary

The present disclosure provides a speech data processing method and apparatus, a device and a storage medium, wherein the method comprises: a client obtains speech data, uses a speech wakeup engine to perform wakeup recognition for the obtained speech data, and stores the obtained speech data in a cyclic buffer queue; the client obtains a wakeup recognition result returned by the speech wakeup engine upon completion of the wakeup recognition, and if the wakeup recognition result is acknowledged wakeup and it is determined that there occurs a case of oneshot currently, determines an interception starting position and obtains buffer data in the cyclic buffer queue from the interception staring position to an end of the queue, the buffer data at least including the content to be recognized; the client sends the buffer data to a speech recognition engine so that the speech recognition engine performs speech recognition for the content to be recognized in the buffer data. The technical solution of the present disclosure can be applied to improve accuracy of the speech recognition result and the like.