Digital audio coding has been receiving a lot of attention in recent years due to the abundance of processing power and the rapid expansion of the internet.  Specifically, lossy but perceptually lossless audio coding has received most of the attention.  Perceptually lossless coding is the manipulation of the audio signal such that we reduce the overall bandwidth while perceptually the signal remains unchanged.

The benefits of perceptually lossless coding are clear.  Compression ratios of 24:1 are capable using these types of systems whereas other lossless compression methods typically achieve compression ratios of at most 4:1.

Mp3 coding is one type of these perceptually lossless coding systems.  The mp3 system was put forth by the Fraunhaufer institute and was standardized by the Motion Picture Experts Group (MPEG).  The mp3 compression system attains high compression ratios while remaining perceptually lossless by using psycho-acoustic modeling.  Psycho-acoustic modeling is an attempt to mimic how the human auditory system perceives an incoming signal.  If we know, by computing this psycho-acoustic model, that part of an incoming waveform will not be perceived, then we can rid the signal of the unintelligible part without audible distortion.  Thus, the system achieves compression.

Since psycho-acoustic modeling is an attempt to mimic how we hear, one might correctly guess that many experimental tests must be performed on a great number of people before an appropriate universal model can be adopted.  All psycho-acoustic models have been formulated by vast experimentation on the human auditory system.  In this experimentation, a couple of the important discoveries were made in the topic of masking.

There are two types of masking:  frequency masking and temporal masking. 
Frequency masking can best be described by viewing Figure 1.  This figure depicts how a strong tonal signal can drown out weaker, nearby signals in the human auditory system.  It should be noted that frequency masking occurs mostly within the same critical band and is considered to be instantaneous.

There is another type of masking, called temporal masking, where signals are irrelevant within a certain time period of a large amplitude signal.  Temporal masking is also restrained to the same critical band as the large amplitude signal.

It is these types of masking that are mainly used to determine accurate psycho-acoustic models in mp3 compression.