Test 1 in AudioInputProcessorIntegrationTest.cpp

[shikha-bansal]

I have the following questions regarding the tests in the AudioInputProcessorIntegrationTest.cpp file:

1. class wakeWordTrigger , use TAP as initiator
   original : m_aip->recognize(audioProvider, Initiator::TAP, aipBegin, aipEnd, keyword);
   expected : m_aip->recognize(audioProvider, Initiator::WAKEWORD, aipBegin, aipEnd, keyword);

2. Once I update wakeWordTrigger class to use WAKEWORD initator , the test works fine if ALEXA is the wakeword , but if a different wakeword is used AVS sends back an exception instead of directives.
   Diff keyword :
   Context built successfully:{"context":[{"header":{"namespace":"SpeechRecognizer","name":"RecognizerState"},"payload":{"wakeword":"XYZ"}},{"header":{"namespace":"Speaker","name":"VolumeState"},"payload":{"volume":50,"muted":0}}]}
   ALEXA keyword :
   Context built successfully:{"context":[{"header":{"namespace":"SpeechRecognizer","name":"RecognizerState"},"payload":{"wakeword":"ALEXA"}},{"header":{"namespace":"Speaker","name":"VolumeState"},"payload":{"volume":50,"muted":0}}]}

The only difference in the 2 context is the keyword.

Any suggestions ? Is the current release of AVS meant to work only with ALEXA keyword ?

[kencecka]

Hi Shikha,

You are correct - AVS only supports "ALEXA" as a keyword. See the following link for specific documentation on this:

https://developer.amazon.com/public/solutions/alexa/alexa-voice-service/reference/context#recognizerstate

Let me know if you have further questions.

Ken

[shikha-bansal]

Hi Ken ,
Thanks for your response. Wrt to point 1 ( wakeword vs tap as initiator).

A ) Wakeword as Initiator
Context built successfully:{"context":[{"header":{"namespace":"SpeechRecognizer","name":"RecognizerState"},"payload":{"wakeword":"ALEXA"}},{"header":{"namespace":"Speaker","name":"VolumeState"},"payload":{"volume":50,"muted":0}}]}

B) Tap as initaitor
Context built successfully:{"context":[{"header":{"namespace":"SpeechRecognizer","name":"RecognizerState"},"payload":{"wakeword":""}},{"header":{"namespace":"Speaker","name":"VolumeState"},"payload":{"volume":50,"muted":0}}]}

A , fails the test with the following error :
home/shikha.bansal/AVS/0.3/alexa-client-sdk-master/Integration/test/AudioInputProcessorIntegrationTest.cpp:967: Failure
Expected: (params.type) != (TestDirectiveHandler::DirectiveParams::Type::TIMEOUT), actual: 4-byte object <06-00 00-00> vs 4-byte object <06-00 00-00>

Please suggest.

[kencecka]

Hi Shikha,

This is probably failing because you are missing the start/end indices in the Recognize Event payload:

        "payload": {
          "wakeWordIndices": {
            "startIndexInSamples": {{LONG}},
            "endIndexInSamples": {{LONG}}
          }   
        }

The sample wakeword detector included in release 0.3 is not able to detect these endpoints for the keyword, so they are omitted. AVS expects to see these indices with a "WAKEWORD" initiator. The integration test sets the initiator to "TAP" in this case, as there is really no difference between a wakeword and a tap if no indices are available.

Other wakeword detectors do provide the start/end indices, and for those, you can specify a "WAKEWORD" initiator, and see the event handled correctly by AVS.

Ken

[shikha-bansal]

Thank you for your explanation Ken.

Best
Shikha

[JamieMeyers]

Hi @shikha-bansal,

Are you still seeing the test failure, or are your issues resolved?

Thanks,
Jamie

[shikha-bansal]

Resolved. Thanks for your help.