Voice chat using Sound API?

Question

Hi there,
I'm a university student, trying to find a way to create an application in java with voice chat feature. It will actually be a call center simulator app consisting of two parts, a server and a client part:
The multithreaded server manages the clients, and the JPA database;
The client's job is to login to the server, and then he'll get a randomly selected person's data and phone number by the server - in this case, IP address is the phone number, and the person is another available client. Then he can call that person, ask the questions regarding for example, a bank card (if it is a bank's call center), jot down the answers to a form, and then hang up; then the client will send said person's data to the server, who updates it in the database.
For the voice chat part, I want to use the Java Sound API, and I've found this example:
[Voice Chat Using Java|http://javasolution.blogspot.com/2007/04/voice-chat-using-java.html]
But although this example works, there's a delay of 1.5 - 2 seconds in the audio.
My questions are:
- Is it a good idea to implement a voice chat feature using the Java Sound API, something similar to the example above on the link?
- Is it possible to somehow fix that delay? if yes, then how?
Also, how would you imagine/structure a program like that?
Let me remind you that it's only a simulation, and we're not talking about actual phone-to-net calls or something like that; only LAN and internet calls, so clients should need no more than a microphone and a speaker to work with it, and also, the server shouldn't need extra hardware,too.
Thank you in advance for your replies,
Ben Dash
university student

Answer 1

bendash wrote:
My questions are:
- Is it a good idea to implement a voice chat feature using the Java Sound API, something similar to the example above on the link?If you need to do the voice chat in Java for whatever reason, then yes...
- Is it possible to somehow fix that delay? if yes, then how?From what I saw from glancing at that code, it's transporting the audio streams via a TCP stream. TCP streams retransmit data when it's lost, which is actually something that you don't want in a real-time application. If you lose some packets with audio data, and those are retransmitted, your receiver is getting progressively more and more behind (because the audio stream has to wait for the missing bytes to arrive before it can continue.
For a simple application like yours, your best bet would be to divide the audio stream into buffer-sized chunks (1024 bytes might be a good buffer size), and transmit those in UDP packets. Make sure to add a packet number to the UDP packets. On the other end, just write some code so that out of order packets are ignored (so if packet 4 arrives after packet 6, for instance, you drop that packet rather than playing it after 6... so you'd get 123(silence)567...