Maybe Siri will one day be better at understanding my Southern accent, although the personal digital assistant has gotten much better at the task. Apple has been granted a patent (number 10,255,907) for “automatic accent detection using acoustic models.”
In the patent filing, Apple notes that conventional speech recognition systems rely on various statistical models trained from very large amounts of data. The so-called acoustic model is typically trained using extensive quantities of manually transcribed speech audio.
The accuracy of speaker-specific speech recognition accuracy is substantially dependent on the extent to which characteristics of the speaker’s voice are represented in the acoustic model training data. Accordingly, a single acoustic model trained on a finite amount of data will perform poorly for speakers whose voice characteristics are not adequately represented in the acoustic model training data.
Here’s Apple’s summary of the invention: “Systems and processes for automatic accent detection are provided. In accordance with one example, a method includes, at an electronic device with one or more processors and memory, receiving a user input, determining a first similarity between a representation of the user input and a first acoustic model of a plurality of acoustic models, and determining a second similarity between the representation of the user input and a second acoustic model of the plurality of acoustic models.
“The method further includes determining whether the first similarity is greater than the second similarity. In accordance with a determination that the first similarity is greater than the second similarity, the first acoustic model may be selected; and in accordance with a determination that the first similarity is not greater than the second similarity, the second acoustic model may be selected.”
Of course, Apple files for — and is granted — lots of patents by the U.S. Patent & Trademark Office. Many are for inventions that never see the light of day. However, you never can tell which ones will materialize in a real product.
