Voice Extraction Unit Noise Removal via Sound Source Direction Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Voice interactive systems face challenges in distinguishing user speech from noise generated by external devices, leading to erroneous processing due to increased processing load and time, as existing methods require complex audio recognition and subtraction of common data from multiple signals.
Innovation Solution
An information processing device and method that analyze sound source directions and frequency characteristics of external apparatus output sounds, registering these in a database to remove noise from user speech inputs, thereby isolating clear user speech for accurate recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complex audio recognition processing is performed on multiple audio signals to distinguish user speech from noise, then the accuracy of speech recognition is improved, but the processing load and processing time increase
Solution Approach 1:
The system performs preliminary audio recognition processing on a reference audio signal (from the reference microphone) to identify and extract noise components before the main speech recognition process. This pre-processing step creates a noise profile that is then used to filter the target audio signal, reducing the complexity of the main recognition process while maintaining accuracy.
Solution Approach 2:
The invention extracts noise components from the audio signal by comparing the target audio (from target microphone) with reference audio (from reference microphone). The noise portion is identified and separated from the target speech signal, allowing clean speech recognition without complex processing of the entire mixed signal.
2Measurement precision
If multiple audio signals are processed individually to remove noise, then the quality of user speech extraction is improved, but the processing cost increases
Solution Approach 1:
The system combines multiple audio signals (target audio from target microphone and reference audio from reference microphone) into a single processed output. By merging these signals and performing unified noise cancellation processing, the system achieves high-quality speech extraction without the need for separate independent processing of each signal, thereby reducing processing cost.
3Reliability
If the system processes all input sounds as user speech, then no speech is lost, but erroneous processing increases
Solution Approach 1:
The reference microphone acts as an intermediary that captures ambient noise and device output sounds separately from the target speech. This reference signal serves as a mediator to identify and remove noise components from the target audio, allowing the system to reliably distinguish between actual user speech and noise without erroneous processing.
Data Source
AI summary
Provided is a device that includes a user spoken voice extraction unit that extracts a user spoken voice from a microphone input sound. The user spoken voice extraction unit analyzes a sound source direction of an input sound, determines whether the input sound includes an external apparatus output sound on the basis of sound source directions of external apparatus output sounds recorded in a database, and removes a sound signal corresponding to a feature amount, for example, a frequency characteristic of the external apparatus output sound recorded in the database, from the input sound to extract a user spoken voice from which the external apparatus output sound has been removed upon determining that the input sound includes the external apparatus output sound.