Authors:
Manda Miller, Albert Bennah, Kristen Chung, Ashley Wehr, Akshay Kumar Pandit
Abstract:
A methodology is proposed to understand whether the wearer of headphones with adaptive noise cancellation is singing or speaking to provide a more optimal experience while wearing headphones and listening to music
Background:
Adaptive earbuds have adaptive noise cancellation to adjust noise levels in real-time based on sounds they detect around them. However, these headphones don’t yet understand ‘singing’ vs ‘speaking’ and turn down the volume for any detection of your voice. This prevents earbud wearers from enjoying their headphones while belting out their favorite chorus or tune because the headphones think the user is speaking to someone.
Description:
Adaptive headphones alter the volume of sound coming through the headphones when a speech signal is detected. But speech isn’t singing, and sometimes while singing, you would like the volume to stay stable.
This disclosure poses a way to detect singing vs speech by incorporating a sensor that measures:
- Pitch: Music tends to have certain pitch patterns and logical spacing of pitches. Notes are sung that are spaced higher or lower and pitches are usually spaced by recognizable and oftentimes consistent intervals.
- If the intervals exist in a repeatable pattern, it is likely that the voice making the sound is “singing”
- Word and Pitch Match Detection: Headphones are usually hardwired or connected through Bluetooth to another device producing music. If the headphones are connected and allowing the user to listen to music, the headphones overlay the music from the device with the voice it hears, if pitch and/or words match, then the person is “singing”
TGCS Reference 00175
