Speech Detection Apparatus Using Segmented Energy Differentiators
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current speech detection technologies are impractical for low-power applications like hearing aids due to high computational requirements and latency issues with cloud processing, making it challenging to implement accurate and efficient speech classification.
Innovation Solution
A speech detection apparatus and method that includes a signal conditioning stage to filter and calculate energy values, a detection stage with multiple differentiators to identify speech and noise, and a combination stage to provide indications of speech presence or absence, utilizing modulation-based, frequency-based, and impulse detectors to efficiently classify speech in audio signals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If cloud-based processing is used for speech detection, then speech detection accuracy can be improved, but latency increases and it becomes impractical for low-power applications
Solution Approach 1:
The speech detection system is segmented into multiple independent differentiators (modulation-based, frequency-based, impulse detector) that operate in parallel. Each differentiator processes specific acoustic features independently and provides speech-detection indications that are combined to make the final speech presence determination, enabling accurate local processing without cloud dependency
Solution Approach 2:
The patent introduces intermediate energy value sequences and speech-detection indication signals as mediators between the raw acoustic signal and the final speech classification. The signal conditioning stage computes energy values that serve as intermediaries, which are then processed by multiple differentiators to produce indication signals that are combined in the combination stage, enabling efficient local decision-making
2Measurement precision
If complex speech detection algorithms are implemented, then speech detection accuracy is improved, but computational power requirements increase
Solution Approach 1:
The detection algorithm is segmented into multiple simple differentiators (modulation-based, frequency-based, impulse detector) that each perform computationally lightweight operations. Instead of implementing one complex algorithm, the system uses several simple algorithms that operate in parallel with minimal computational overhead each, reducing overall power consumption while maintaining accuracy
Solution Approach 2:
The patent transforms the speech detection problem by changing parameters from analyzing raw acoustic signals directly to analyzing computed energy value sequences. By pre-processing the signal to extract energy values in different frequency bands and temporal modulations, the system simplifies the detection task and reduces computational complexity of the differentiators
3Power
If simple speech detection methods are used, then computational resources are minimized, but speech detection accuracy deteriorates
Solution Approach 1:
The patent merges the outputs of multiple differentiators (modulation-based, frequency-based, impulse detector) in the combination stage to make the final speech presence determination. By combining the speech-detection indication signals from different simple detectors, the system achieves accuracy comparable to complex algorithms while keeping individual detector complexity low
Solution Approach 2:
The speech detection apparatus is designed with multi-functional differentiators that analyze multiple acoustic features (temporal modulation, frequency distribution, impulse characteristics) using unified energy value sequences. This universal approach allows simple computational operations to serve multiple detection purposes simultaneously, improving accuracy without proportionally increasing computational resources
Data Source
AI summary
In a general aspect, an apparatus for detecting speech can include a signal conditioning stage that receives a signal corresponding with acoustic energy, filters the received signal to produce a speech-band signal, calculates a first sequence of energy values for the received signal and calculates a second sequence of energy values for the speech-band signal. The apparatus can also include a detection stage including a plurality of speech and noise differentiators. The detection stage can being configured to receive the first and second sequences of energy values and, based on the first sequence of energy values and the second sequence of energy values, provide, for each speech and noise differentiator of the plurality of speech and noise differentiators, a respective speech-detection indication signal. The apparatus can also include a combination stage configured to combine the respective speech-detection indication signals and based on the combination of the respective speech-detection indication signals, provide an indication of one of presence of speech in the received signal and absence of speech in the received signal.