CODEWITHSUNDEEP
javaaudiowavtwos-complementjavax.sound.sampled
Ryan Foster
Ryan Foster

Reputation: 101

What does the raw data in a .wav file mean?

So I'm trying to extract the amplitudes of different samples from a .wav file to be able to reconstruct the waveform with regards to time. While researching the format of a .wav file, I found some very useful information. However, I was confused when it came to the raw data portion. The article I found stated,"In the case of 16 bit PCM, 16-bit samples are stored as 2's-complement unsigned integers, ranging from 0 to 65535". I'm confused because I don't know how the file distinguishes between positive and negative amplitudes if it's unsigned. For example, if one amplitude is 5 represented as 00000000 00000101 in binary, then using the 2's complement -5 should be represented as 11111111 11111011 which is 65531 in decimal as an integer. This makes me think an amplitude of -5 is perceived as MUCH larger than a +5 amplitude. Clearly this isn't how it works so what am I doing wrong? Also, when I save my .wav file into a binary array and output the values, I get negative numbers! Where do these negative numbers come from if the article said only values between 0 to 65535 should be stored. Just very confused, really need to have things cleared up. The link of the article I was referring to is attached down below. The quote of what the author said is found at the very bottom of the page. Thanks.

http://www.ievs.ch/projects/var/upload/Documentation%20Microsoft%20Wave%20File%20Format.pdf

Upvotes: 0

Views: 1317

Answers (1)

Phil Freihofner
Phil Freihofner

Reputation: 7910

I recommend bring the .wav data into Java via AudioSystem methods, rather than attempting to work on the raw bytes. Once brought in, making use of the appropriate AudioFormat, there is no need to deal with header or any sort of packet-organizing info for the data.

When I've brought in data this way, the decoded PCM is stored as signed shorts. Thus it ranges from -32768 to 32767. If 16-bit encoding, the values are obtained from two bytes, assembled in an order specified by the bigEndian flag.

https://docs.oracle.com/javase/9/docs/api/javax/sound/sampled/AudioFormat.html

A second concern is that the individual PCM values probably should not be thought of as a useful indication of volume. Instead volume is usually determined by doing a RMS (root-mean-square) analysis of the signal.

The RMS over all time of a periodic function is equal to the RMS of one period of the function. The RMS value of a continuous function or signal can be approximated by taking the RMS of a sample consisting of equally spaced observations.

Upvotes: 0

Related Questions