Multi-step validation of wakeup phrase processing
US9972343B1 · kind B1 · utility
Assignee
Inventors
Key dates
| Filing date | Jan 8, 2018 |
| Grant date | May 15, 2018 |
| Priority date | — |
| Expiry date | Jan 8, 2038 |
Classification
- Technology area (CPC G)Physics
- CPC primaryG10L2025/783
- WIPO fieldComputer technology
- WIPO sectorElectrical engineering
Abstract
Various embodiments describe a multi-step process for determining whether the keyword has been spoken. The first tier uses an onboard digital signal processor (DSP) to process converted audio signals. If the DSP analysis cannot confirm the keyword due to noise in the signal, a second tier of processing may use an applications processor (AP) and a more stringent audio filter to try to verify the keyword. If the AP analysis cannot confirm the keyword, a third tier of processing may use server side processor and an even more stringent audio filter to try to verify the keyword. Certain signal to noise ratios (SNRs) of the detected and converted audio signal are used to decide when and if to try the second and third tier of processing.
Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.