question

nathanmani asked Nov 25, '15

AVS response is not expected one when more silence at end

The AVS response is varying based on silence duration after speech end for same audio/speech data. Is it expected behaviour. If so, we have to cut the silence. Is there any limitation of silence duration before speech start and after speech end.

alexa voice service

10 |5000

Attachments: Up to 2 attachments (including images) can be used with a maximum of 512.0 KiB each and 1.0 MiB total.

Eric@Amazon answered Nov 25, '15

No, extra silence is not expected behavior. Can you send us request ids? You can find them by printing out the response headers. That will help us diagnose this problem.

10 |5000

Attachments: Up to 2 attachments (including images) can be used with a maximum of 512.0 KiB each and 1.0 MiB total.

nathanmani answered Dec 1, '15

Are you referencing amazon account id (request ids)?

10 |5000

Attachments: Up to 2 attachments (including images) can be used with a maximum of 512.0 KiB each and 1.0 MiB total.

nathanmani answered Dec 1, '15

Hi Eric, Also, the response from AVS is varying for time to time..ie. feeding same audio but getting different response.. What would be the issue?

10 |5000

Attachments: Up to 2 attachments (including images) can be used with a maximum of 512.0 KiB each and 1.0 MiB total.

Eric@Amazon answered Dec 2, '15

I'm referring to the request IDs. When you get a response back from AVS, the HTTP header information will have a header with "x-amzn-requestid: 123b76fffeb7c361-00000264-00000641-17ff9ec37c763ce0-5795db7b-21" (but with some other request ID). We can use that to help debug what's going on. Additionally, what audio are you sending to AVS? Can you send that to us as well? Thanks, Eric

10 |5000

Attachments: Up to 2 attachments (including images) can be used with a maximum of 512.0 KiB each and 1.0 MiB total.

nathanmani answered Dec 2, '15

Here is my Request ids: 65bbf509-98d9-11e5-8de1-17fc865555ec 8df978c6-98d9-11e5-a1b8-13169954681a Both gave different response for same audio input Shall i send the audio file to Eric@Amazon

10 |5000

Attachments: Up to 2 attachments (including images) can be used with a maximum of 512.0 KiB each and 1.0 MiB total.

Eric@Amazon answered Dec 2, '15

Instead of emailing it to me, can you post the audio file to a publicly accessible place and a link to it here? Thanks, Eric

10 |5000

Attachments: Up to 2 attachments (including images) can be used with a maximum of 512.0 KiB each and 1.0 MiB total.

nathanmani answered Dec 3, '15

I have shared two audio files avsip1.bin avsip2.bin (RAW PCM files). Please feed them one by one and do it for multiple iterations. https://www.dropbox.com/sh/om25bjb6opj90i5/AADPBvWO8XKhtOZUXJq9cOsEa?dl=0

10 |5000

Attachments: Up to 2 attachments (including images) can be used with a maximum of 512.0 KiB each and 1.0 MiB total.

Eric@Amazon answered Dec 3, '15

Thanks for sending those! They'll be very helpful as we try to improve Alexa. I was able to duplicate the problem, and I've forwarded the information to the right people. The problem seems to be that Alexa didn't quite understand what was being said. If you go to https://alexa.amazon.com, you should be able to see what Alexa heard (click 'more' on an information card). For now, you can try other variations on your question until you get a more consistent response. The silence at the end of the audio file shouldn't make a difference. Thanks again for bringing this to our attention!

10 |5000

Attachments: Up to 2 attachments (including images) can be used with a maximum of 512.0 KiB each and 1.0 MiB total.

nathanmani answered Dec 4, '15

Thank you so much.... We are not using Amazon Echo device.We have ported AVS on Linux PC and testing the performance. Still we are able to look what Alexa heard in " https://alexa.amazon.com" ?. If so how? Please

10 |5000

Attachments: Up to 2 attachments (including images) can be used with a maximum of 512.0 KiB each and 1.0 MiB total.

nathanmani answered Dec 4, '15

Also, whether the audio is fine for AVS (means power level, frequency response etc)?

10 |5000

Attachments: Up to 2 attachments (including images) can be used with a maximum of 512.0 KiB each and 1.0 MiB total.