18172711. SPEECH ENHANCEMENT USING PREDICTED NOISE simplified abstract (QUALCOMM Incorporated)
Contents
SPEECH ENHANCEMENT USING PREDICTED NOISE
Organization Name
Inventor(s)
Asif Mohammad of San Diego CA (US)
Fatemeh Alishahi of Los Angeles CA (US)
Maliha Arif Zain of Orlando FL (US)
Sajeer Fazil of San Diego CA (US)
SPEECH ENHANCEMENT USING PREDICTED NOISE - A simplified explanation of the abstract
This abstract first appeared for US patent application 18172711 titled 'SPEECH ENHANCEMENT USING PREDICTED NOISE
Simplified Explanation: The patent application describes a device that uses processors to predict and subtract noise from an input audio signal containing speech.
Key Features and Innovation:
- Processors obtain an input audio signal with speech.
- Predicted noise signal is generated based on a trained model.
- Predicted noise signal is subtracted from the input audio signal.
- Output audio signal is generated without the predicted noise.
Potential Applications: This technology could be used in:
- Speech enhancement systems
- Noise cancellation devices
- Audio editing software
Problems Solved:
- Eliminates background noise in audio recordings
- Improves speech clarity in noisy environments
Benefits:
- Enhanced audio quality
- Improved speech intelligibility
- Better user experience in audio applications
Commercial Applications: Potential commercial uses include:
- Audio recording equipment
- Communication devices
- Entertainment systems
Prior Art: Readers can explore prior art related to noise cancellation and speech enhancement technologies in the field of audio signal processing.
Frequently Updated Research: Stay informed on the latest advancements in noise reduction and speech enhancement technologies for audio applications.
Questions about Noise Cancellation Technology: 1. How does the device predict noise in the input audio signal? 2. What are the potential limitations of subtracting predicted noise from the input audio signal?
2. Another relevant generic question, with a detailed answer.
Original Abstract Submitted
A device includes one or more processors configured to obtain an input audio signal including at least first speech of a first person. The one or more processors are configured to generate a predicted noise signal based on processing of the input audio signal by a trained model. The one or more processors are configured to subtract the predicted noise signal from the input audio signal to generate an output audio signal.
