Speech Recognition Guidance Speech Interference Removal
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional speech recognition systems require an additional microphone to remove guidance speech, leading to high costs and complexity, and result in a decline in recognition rates when guidance speech is outputted simultaneously with recognition speech.
Innovation Solution
A speech recognition apparatus that uses the guidance speech-data as a reference to remove the guidance speech from the recognition speech-data, eliminating the need for an additional microphone by converting and synchronizing the sampling rates of guidance and reference speech-data to match the recognition speech-data, thereby preventing recognition rate decline.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If an additional microphone is used to remove guidance speech, then the recognition rate is maintained, but the device complexity and cost increase
Solution Approach 1:
The patent creates a reference speech signal by copying the guidance speech output and processing it to match the characteristics of the actual guidance speech picked up by the microphone. This virtual reference signal is then used for subtraction to remove guidance speech interference, eliminating the need for an additional reference microphone.
Solution Approach 2:
The patent introduces a reference speech generation unit that acts as an intermediary between the guidance speech output and the speech recognition process. This unit synthesizes a reference signal that mediates the subtraction operation, allowing guidance speech removal without requiring a separate physical microphone.
2Reliability
If an additional microphone is used to remove guidance speech, then the recognition rate is maintained, but the system cost increases
Solution Approach 1:
The patent creates a reference speech signal by copying the guidance speech output and processing it to match the characteristics of the actual guidance speech picked up by the microphone. This virtual reference signal is then used for subtraction to remove guidance speech interference, eliminating the need for an additional reference microphone.
Solution Approach 2:
The system uses its own guidance speech output as the basis for creating the reference signal, making the system self-sufficient. The guidance speech that is already being outputted is reused and processed to generate the reference signal needed for interference removal, eliminating external hardware requirements.
3Ease of operation
If guidance speech is outputted simultaneously with recognition speech, then user guidance is provided, but the recognition rate decreases due to interference
Solution Approach 1:
The patent extracts the guidance speech component from the mixed speech signal by subtracting the generated reference speech signal from the microphone input. This separation isolates the user's recognition speech from the guidance speech interference, enabling accurate recognition even during simultaneous output.
Solution Approach 2:
The patent converts the harmful effect of guidance speech interference into a beneficial process. By using the guidance speech output itself to generate the reference signal for subtraction, the system transforms the interfering signal into a tool for its own removal, maintaining recognition accuracy during simultaneous operation.
Data Source
AI summary
In a speech recognition apparatus, a speech driver fetches a guidance speech-data as a reference speech-data, and outputs the reference speech-data to a recognition core unit. A guidance speech into which the guidance speech-data is converted is outputted by a speaker to cause a microphone to receive the outputted guidance speech, which will be converted into an inputted guidance speech-data. Even in such case, a speech recognition engine removes the inputted guidance speech-data by using, as the reference speech-data, the guidance speech-data that is before being converted into the outputted guidance speech.