Lossy compression format at a glance
“As you know, the music we listen to consists of a set of signals, each of which has its own characteristics, including loudness.
The human hearing aid is designed so that we do not distinguish or poorly distinguish a weak (low) signal from the background of a strong (strong) signal. This principle forms the basis of modern means of compression (compression) of audio data.
If we imagine that a signal of a certain length is divided into many parts, and each part is processed in such a way that a weaker signal, which is difficult to distinguish from the background of a strong one, falls under the knife, and one remains a signal louder, then this will be an approximate audio compression model. Consequently, the level of data compression will depend on how many parts (samples) the original file will be divided into and how many weak signals from each individual sample will be removed (what the bit rate will be, the number of bits in a sample). sample of a specified duration). This coding principle is called lossy coding or lossy coding.
Ogg Vorbis is a completely open and patent-free audio format that allows you to store and transmit audio information with high sound quality (44.1-48.0 kHz sample rate, more than 16 bits, polyphony (multi-channel audio) ) and bit rates ranging from 16 to 512 Kbps per channel. In this case, the number of processed channels can reach 255.
MP3 or MPEG-1 Layer 3 audio is by far the most popular format for storing and transmitting compressed data. This format was developed by the Fraunhofer Institut, Germany. “Http://ru.wikibooks.org/wiki/Compression_Audio_data_with_lossy
Comparative tests
Sound Forge 7.0 (Spectral Analysis / Spectrum Analysis function) was used for the analysis of the sound signal.
“Spectral analysis is a signal processing technique that can reveal the frequency content of a signal. Solving the problems of spectral analysis is possible through the use of the fast Fourier transform, which makes it possible to determine the contribution of individual components of the vibration spectrum to the overall vibration picture. “Http: //masters.donntu. edu.ua/2007/fema/belinskaya/library/a4/art4.htm
The following graphs were obtained in the form of an amplitude distribution in the frequency domain, the spectrum of the signal is presented using a Blackman-Harris / Blackman-Harris window and a maximum sampling frequency (FFT size) of 65536, this gives allows you to analyze the smallest details of the signal at frequencies around 20,000 Hz, without smoothing.
The analysis of the spectrum of the compressed signal assumes the presence of a recording of the original quality, for this we use a licensed audio CD made in the USA “Kevin Yost – Bongo Madness”, with standard characteristics 44100 Hz / 16 bit
The rich electronic sound spans the entire frequency spectrum and captures even the inaudible range (20,000 Hz to 22,000 Hz), as can be seen in the graph below. Considering that it is generally possible to notice codec compression at higher frequencies, the 10-20 kHz range will be considered.
