similar to: speex voice seems to be bit breaking over long distance.

Hi, Ok, what you suggest sound logical to me. Currently, I have done a small trick to prevent this problem. What I did is that whenever windows request a voice packet from me and if I do not have the voice packet, I repeat the previous packet. Hence, all the breaking portion is filled with previous packet. This trick seems to work so far. I am not sure what is the side effect. I think jitter

> Ok, what you suggest sound logical to me. Currently, I > have done a small trick to prevent this problem. What > I did is that whenever windows request a voice packet > from me and if I do not have the voice packet, I > repeat the previous packet. Hence, all the breaking > portion is filled with previous packet. This trick > seems to work so far. I am not sure what is the

Hi Jean, > Actually, Speex has Packet Loss Concealment (PLC) > builtin. If a packet > is missing, instead of repeating the previous one, > just try decoding by > passing NULL instead of the SpeexBits struct. > Thanks, I have made the above changes and the effect seems to be better now. > > I think jitter buffering is more correct way to > solve > > this problem

I use Speex with dialup modem users. Even if the dialup modem is "56K", you should assume that upstream bandwidth available is only 20-30kbps at the most. I use 16kHz wideband mode and VBR quality 2. Also, I send 80ms (4 frames) per packet, because there is an overhead of approximately 33 bytes per packet due to UDP (8 bytes), IP (20 bytes), and PPP (~5 bytes) headers. If you

Hi, Have anyone using speex experience the following effect: When I say "hello", it will be become "hellololololololo....". This does not always happened. The chance of happen is quite random. This effect does not happen in my LAN. It only happen when ping time is about 500 milisecond or more. Is it a speex problem? Any suggestion how to resolve this issue? Thanks in

Hmmm... sorry i mislead you Tay... i didn't realise my encoder (libfishsound) was again shifting the data i gave it back to being 32767 based internally. Zen. ----- Original Message ----- From: "Jean-Marc Valin" <Jean-Marc.Valin@USherbrooke.ca> To: "Tay YueWeng" <yueweng@yahoo.com> Cc: "speex" <speex-dev@xiph.org> Sent: Wednesday, December

Hi Jean, > > > Well, just trace it, how many times are you > calling > > > speex_decode_int()? > > > > Maximum is 51 times per second. Will this cause > any > > CPU high utilization? > > That's normal... What CPU are you using? If it's a > fixed-point CPU, then > the reason is probably just the fact that the packet > loss

Hi, This applies to everyone having (or suspecting) problems with Speex. The first thing to do is to encode the file in wav format and use speexenc/speexdec on it. If you're getting something different with your application, it's likely buggy. If the result isn't OK, then it can be: 1) Normal given the bandwidth/bit-rate used 2) A conditioning problem with your audio (i.e. DC not

Dear all, I have one simple question. I understand that speex_encode and speex_decode takes float * as an arguement to encode and decode the sound. However, when I get the PCM data from the sound card under win32, it is a just 16 bit array. May I know how do I convert this 16 bit value to speex float format and to convert back? Is there got any routine to do this? YueWeng

Hi all, I have deploy speex 1.1.6 in my application. With no option set, I can hear that the voice sounds a little bit artificial like robot. Any idea what causes this? I use openh323 with speex, but it seems ok. Is it neccessary for me to use more other filter prior to encode the sound or after decode my sound? yueweng __________________________________ Do you Yahoo!? Yahoo! Mail -

Dear all, First, my aim is to achieve VoIP using VBR and DTX under Win32. I face a problem using speex 1.0.4 and need some help. My voice is ok and no background noise when I do NOT normalize 16 bit value to floating value. Normalized means dividing the 16 bit value by 32767. Turning on VBR is also ok but DTX has no effect. However, the speak is has a continous background beep sound AFTER I

Dear all, I am using speex 1.17 at this moment, everything works great. However, I face a problem when no packet arrived from network for a few second, my CPU usage is 100%. I step though my code and seems that (not confirmed) the speaker callback WaveOutCallback() function which call speex_decode_int(decoder_state, NULL, shortData) (when no data arrived for PLC purpose) seems to consume a

Hi, > 1) Normal given the bandwidth/bit-rate used Do you mean the bit-rate that I should set in the speex codec? > 2) A conditioning problem with your audio (i.e. DC > not removed) What is DC? YueWeng __________________________________ Do you Yahoo!? Yahoo! Mail - now with 250MB free storage. Learn more. http://info.mail.yahoo.com/mail_250

On Fri, Nov 14, 2008 at 3:57 AM, Vincent Burel <vincent.burel at vb-audio.com>wrote: > > Speech compression algorithms always are tunned to particular freq, > > else they would take tons of time. That's because they use knowledge > > that speech pitch (and other params) lies in well specified regions. > > Thus if you feed algorithm with wrong freq, you'll

Hello Mailing List, I am a Speex supporter and user that would really like to know how much memory Speex uses to decode a 8kHz, 16kHz and 32kHz (primarily the 8kHz) and is it possible to use a 1kBytes of RAM to decode a 8kHz stream? (I was thinking of the possibility of using a ATmega168 to decode Speex) //P?r, Sweden

That's pretty bad. Both DirectSoundCapture and WinMM are capable of recording at 16kHz. I don't know why OpenAL would be incapable of handling it. It's not like it's at all rare or new. I would try 16000 and see if it works. Maybe the docs are wrong? Note that one option to retain high quality is to capture at a higher rate and then downsample using a resampling

Hi Randy, One thing I would note is that speex is designed for 8kHz, 16kHz or 32kHz. 160 samples is equal to 20ms of 8kHz audio. Have you tried resampling from 11.025kHz to 8kHz and then using the speex 8kHz mode? (nb, or narrow band). Or, if you want to preserve the higher quality of your 11.025kHz sample rate, resample to 16kHz and use the wideband speex encoder. I believe there is new

Hi, I am using the speex 1.0.4 library from Windows. I have posted my problem before but didn't get a solution. I am doing an VOIP project in which i am recording sound and streaming it to the peer. I wanted to encode and decode wav files that brought me to this site. I am recording sound in the following format:- m_WaveFormatEx.wFormatTag = WAVE_FORMAT_PCM;

> 16 bit ints have a range of -32768 to 32767. If you divide > -32768 by 32767.0 you end up with -1.00003051850948 which > is a bad thing. > > Try normalizing with a value of 32768.0. No. Speex expects values in the +-32767 range, not +-1.0. Just converting from int16 to float *is* the right thing to do. Jean-Marc -- Jean-Marc Valin <Jean-Marc.Valin@USherbrooke.ca>

It seems that I only have the following values available for sampling from the mic. "The value must be 8000, 11025, 22050, 32000, 44100, or 48000" Which leaves 8000 and 32000 for use with speex. I think since this is a game and not a voice application, I'm stuck using the 8kHz rate. What speex setting would you recommend I use for the best quality/performace, what frame size