Azure bing cognitive services speech to text in javascript via REST fails - bing

Question

In JavaScript we use recorder.js to capture microphone input, down sample it to 16kHz, encode it as a WAV file and get a blob.
Next, we obtain the raw blob bytes via a FileReader onload() callback and then use an XMLHttpRequest to send() the raw bytes to Bing.
The XMLHttpRequest includes the headers:
'Ocp-Apim-Subscription-Key' : 'xxxxxx'
'Content-Type' : 'audio/wav; codec=audio/pcm; samplerate=16000'
A sample blob size is 62456 bytes.
FireFox network tracing shows 2 interactions. The first is
Request URL: https://speech.platform.bing.com/speech/recognition/interactive/cognitiveservices/v1?language=en-US&format=simple
Request Method: OPTIONS
and the second
Request URL:https://speech.platform.bing.com/speech/recognition/interactive/cognitiveservices/v1?language=en-US&format=simple
Request Method: POST
content-length: 94476
However, I keep getting the following reply
{"RecognitionStatus":"InitialSilenceTimeout","Offset":29000000,"Duration":0}
FWIW, any idea why the source blob size of 62456 would result in content-length: 94476?
The same raw blob bytes are processed by Amazon Lex properly.
Is there any JavaScript RESTful example?
Many thanks.
/--------------------------------------------------------------
After putting together the test case below I also tried the following without success.
console.log("Send to BING blob");
var self = this;
console.log(blob);
var msUrl = 'https://speech.platform.bing.com/speech/recognition/interactive/cognitiveservices/v1';
msUrl += '?language=en-US';
msUrl += '&format=simple';
console.log(msUrl);
var xhr = new XMLHttpRequest();
xhr.onload = function(evt) { console.log('onload', xhr, evt);};
xhr.open('POST', msUrl, true);
xhr.setRequestHeader('Accept', 'application/json;text/xml');
xhr.setRequestHeader('Ocp-Apim-Subscription-Key', 'xxx');
var bingContentType = 'audio/wav; codec=audio/pcm; samplerate=16000';
xhr.setRequestHeader('Content-Type', bingContentType);
xhr.send(blob);

Answer 1

The shorter code version of sending to Bing was fine. The probelm was that the recorder worker's encodeWAV(samples) function did not
take into account the down sampling to 16000. The function was incorrectly writing the captured sampling rate as the header value. The lines to be tweaked are:
view.setUint32(24, downSampleRate, true);
view.setUint32(28, downSampleRate * 2, true); /*MONO*/
Apparently AWS Lex ignores the header values as it only expects 16kHz mono whereas the Bing service has to look at the header information to determine which of the audio formats supported is being sent.

Answer 2

Today I came across this problem, after spending half an hour, I was able to find the real cause of my issue. Let me go through the steps which are mentioned in this link.
Verified my Bing speech API is in running status.
Verified my key by running the below code in
$FetchTokenHeader = #{
'Content-type'='application/x-www-form-urlencoded';
'Content-Length'= '0';
'Ocp-Apim-Subscription-Key' = ''
}
$OAuthToken = Invoke-RestMethod -Method POST -Uri https://api.cognitive.microsoft.com/sts/v1.0/issueToken -Headers $FetchTokenHeader
show the token received
$OAuthToken
As mentioned in the last point in that link, InitialSilenceTimeout may be the result of the unformatted/invalid wav file. So I downloaded a new wav file from internet and tested with it.
Bingo, that worked. And finally, I was able to get my speech in text format