Topic: Microphone, Voice Encoding

As most of us know, the new microphone has, up to this point, been implemented with the following...

-- Push and hold button to start the recording from the mic.
-- Release the button to signal you are done recording.
-- The stored audio file (mp3 encoded?) is then sent back to the server
-- On the server, some decoding of speech to text is employed to extract the words spoken
-- The proper service or event is triggered (such as reading the weather, playing a radio station, etc).

So, a couple of questions I have that maybe you smart folks know the answer to...

-- What is the audio encoding format?
-- What is the time limit on how long you can speak (memory limitations)?
-- Can the audio upload and event triggered happen on the same connection to the server (I think it does this now)?
-- Can the 'record audio' button be 'pushed' during a server ping, allowing for automatic audio capture without user intervention?

Roy

Re: Microphone, Voice Encoding

Roy, to the best of my knowledge:

>>>>>>>>>>>>-- What is the audio encoding format?
.wav file with IMA ADPCM codec, mono, 8 KhZ sampling rate, audio sample size 4 bits, bit rate 32kbps

You could change the sampling rate by hacking the bootcode, but the ADPCM format will be fixed (it's the only one supported for recording by the VS103b chip)

>>>>>>>>>>>>>>><-- What is the time limit on how long you can speak (memory limitations)?
Current bootcode limits to 8 seconds. I don't know what is the maximum you can patch to without having memory limitations; probably always less than 5 minutes). If you want to patch and test, the offset is 0x8F0B and consists of 4 bytes LSB first with the value in miliseconds.
So currently
40 1F 00 00 -> 00001F40 hexa -> 8000 dec -> 8 seconds (this takes 32 kbytes of memory, and the ram chip in the bunnyhas 1 Mbyte)
If you do the tests please post your results

>>>>>>>>>>>>>>>>-- Can the audio upload and event triggered happen on the same connection to the server (I think it does this now)?
Sorry?

->>>>>>>>>>>>>>>>>>- Can the 'record audio' button be 'pushed' during a server ping, allowing for automatic audio capture without user intervention?
No way that I know of with the current bootcode. It would be technically possible with a modified bootcode. That is, there's no *hardware* need for the button to be pressed.

Last edited by sache (2007-03-15 23:25:48)

Sache, the bunny trying to know himself better

Re: Microphone, Voice Encoding

Thanks Sache. Let me explain this one...

sache wrote:

>>>>>>>>>>>>>>>>-- Can the audio upload and event triggered happen on the same connection to the server (I think it does this now)?
Sorry?

In other words, currently the Nabaztag has a 2 minute polling interval. But, I've noticed that certain commands are processed 'immediately'.

So, when I give the command 'Weather', I get an almost instant response. So, I assume it delivers my voice command and then the response is sent for playback (as opposed to my voice command being uploaded and then the response not showing up until the next polling interval).

Re: Microphone, Voice Encoding

Specifically, bunny first sends the audio in an HTTP POST request. 5 seconds later, it issues a "ping" (HTTP GET p4.jsp) Violet server is suppossed to have processed the audio and have the response ready in these 5 seconds.

Sache, the bunny trying to know himself better

Re: Microphone, Voice Encoding

Interesting. So, that basically suggests that an event (such as recording audio) can change the next ping interval.

Is this the way the button push and ears are done as well? The signal is sent and then a ping is performed a short time later? Or, is this unique to the audio recording?

Re: Microphone, Voice Encoding

Only for audio recording (and RFID too probably). pushbutton and ear movement get directly encoded into the parameters of a ping request

Sache, the bunny trying to know himself better

Re: Microphone, Voice Encoding

HEllo, I have a problem with my Nabaztag mic:

So, my rabbit was all fine, and one day I've turned the volume off (he waked up too early in the weekend!), has I do often, and then, he never spoke again!

This happend in the beginning, when I set it up, but the solution was to connect to a different power source and it went ON, fine.

Now, he does not speak. However, I connect the headphones to him, and I hear his voice, there must be a problem with the speaker, anyone has any ideas? sad

I want my rabbit back! sad