Audio Recognition Engine Segmentation for Low-Bandwidth Mobile Services

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current music recognition systems in mobile services face bottlenecks due to limited bandwidth in GSM networks, leading to poor user experience and potential service failures when handling simultaneous requests, and require users to send large audio clips for identification.

Innovation Solution

A broadcast program content recognition system that splits unrecognized audio samples into sequential portions and appends them to previously recognized samples until recognition is achieved, optimizing the process to function effectively in low-bandwidth environments.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If current music recognition systems are implemented in mobile service context with server farm model, then the system can handle multiple users, but the limited bandwidth of GSM networks causes server load to become too heavy resulting in service failure

Engineering Contradiction:
Improvenumber of users handledVSAvoidservice stability
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent segments the audio recognition process by dividing the audio clip into multiple audio frames and processing them sequentially. The recognition engine processes one frame at a time and can return results after processing only a portion of the total audio data, thereby reducing the amount of data that needs to be transmitted over the network at once and lowering server load during peak usage.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system performs preliminary processing by extracting features from audio frames locally at the mobile device before transmission. This preliminary feature extraction reduces the amount of data that needs to be sent to the server, thereby reducing network bandwidth consumption and server processing load while maintaining recognition accuracy.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If users send audio clips for song identification, then recognition can be achieved, but the audio clips are of considerable size causing network overhead

Engineering Contradiction:
Improverecognition accuracyVSAvoiddata transmission size
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The audio clip is segmented into multiple smaller audio frames that are processed sequentially. The system can achieve recognition results by processing only a portion of the total audio data, allowing users to send smaller data portions while still maintaining recognition accuracy for many cases.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system extracts only the essential features from the audio frames rather than transmitting and processing the complete audio data. This feature extraction process identifies the most relevant characteristics for recognition while discarding redundant information, thereby significantly reducing data transmission size while preserving recognition accuracy.

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP2707975B1System and method for recognizing broadcast program content
Publication Date: 2019.01.02 MAYA INNOVATIONS HOLDINGS PTE LTD
  • EP2707975B1 patent drawingFigure 1
  • EP2707975B1 patent drawingFigure 2
  • EP2707975B1 patent drawingFigure 3

AI summary

A broadcast program content recognition system comprising at least one receiver adapted to sample broadcast program content from a broadcast source; a content recognition engine for recognizing and storing the sampled broadcast program content; and a content database in data communication with the content recognition means; the content database adapted to return information relating to the broadcast program content upon receipt of a query from a client device is disclosed. The system is configured such that in event where the content recognition engine is unable to recognize the sample broadcast program content, the content recognition engine splits the unrecognized sample into at least a first and a second sequential portions and appends the first portion to a previously recognized sample.