Psychoacoustic model and its application in MP3 encoding

Abstract: The psychoacoustic model is the core part of audio perceptual coding, which directly affects the quality and compression ratio of audio coding.

Based on the basic principles of psychoacoustics, the absolute hearing threshold, the masking effect, and the critical frequency band and other related content, combined with the mathematical expression of psychoacoustics, and analyze the algorithm flow in detail according to each module in the standard MP3 encoding program. Finally, the corresponding algorithm is used to describe the pre-echo generation mechanism and its suppression method in MP3 encoding.

Psychoacoustic model and its application in MP3 encoding

Psychoacoustic models and their applications in perceptual audio coding

In this paper, the AAC psychoacoustic model is discussed from the aspects of over-masking, forward and backward time-domain masking, FFT window coefficient analysis, and window shift criterion. Each individually masked…

Auditory psychoacoustic models and their applications in perceptual audio coding

First, the paper describes the masking effect, discusses the principles and algorithms of various perceptual audio coding standards, and focuses on the development process and features of the MPEG audio coding standard. There are two psychoacoustic models: model 1 and model 2. Level 1.

Application of the psychoacoustic model in HDTV digital audio

Digital audio is a very important part of high-definition television (HDTV). In the digital audio codec, the introduction of the psychoacoustic model greatly reduces the complexity of the codec. The basic principles of the psychoacoustic model and various psychoacoustic models in HDTV. ..

Application of the psychoacoustic model in the detection of abnormal loudspeaker sound

China is the largest country in speaker production, and the annual output of speakers can reach hundreds of millions. Due to some unpredictable situations in the speaker design and production process, speaker failure occurs. Therefore, after the speaker is produced, the first product…

Application of the psychoacoustic model in digital audio watermarking

Audio digital watermarking technology plays an increasingly important role in protecting the copyright of digital audio works, and is an effective means of solving the problem of copyright of audio works. digital.A digital watermark algorithm based on a psychoacoustic model is proposed, which can guarantee a good audio signal.. .

An adaptive audio watermarking algorithm that combines the MP3 encoding principle and the psychoacoustic model

Through the investigation of the mp3 encoding algorithm, combined with a psychoacoustic model, a watermarking algorithm for copyright protection of MP3 audio files is proposed. The embedding algorithm runs simultaneously with the compression process. First, reduce the dimension of the embedded watermark information, and then use MP3 encoding to make…

Audio and Video Series: Audio Basics Part 2

Introduction to sound

time domain masking
Masking that occurs between temporarily adjacent sounds.

temporary masking

audio encoding
encoding process
coded process

audio file format
Audio File Format ( wiki ): The format of the file that contains the audio data.

Format Classification
Lossless formats: such as WAV, FLAC, APE, ALAC, WavPack (WV)
Lossy formats: such as MP3, AAC, Ogg Vorbis, Opus
performance comparison
Latency Comparison
delay compare

Efficiency Comparison
efficiency comparison

AAC encoding
AAC (wiki): Advanced Audio Coding, a proprietary audio coding standard for lossy digital audio compression based on MPEG-2, which appeared in 1997.

AAC exhibits better sound quality than MP3 and is intended to replace the MP3 format.

usual format
AAC LC : (low complexity) low complexity specification
AAC HE V1 : (high efficiency) AAC LC + SBR (Spectral Band Replication)
AAC HE V2 : AAC LC + SBR + PS (parametric stereo)
ac profile

data exchange format
ADIF : (Audio Data Interchange Format) Audio Data Interchange Format, can only be decoded from scratch, commonly used in disk files.
ADTS : (Audio Data Transport Stream) audio transport stream format, each frame has a sync word, which can be decoded at any position in the audio stream for data transmission.

Audio and Video Series: Audio Basics

Introduction to sound

Definition: Sound is a sound wave produced by vibration, a wave phenomenon that propagates through a medium (gas, solid, liquid) and can be perceived by the hearing organs of humans or animals.

Essential: Sound is a mechanical wave.

Three elements of sound.
Tone : sound frequency (audio), boys > girls > boys
Volume: The amplitude (amplitude) of the vibration, also known as the pitch
Timbre: The waveform of the sound, which is essentially harmonic, also known as fret, has a lot to do with the material
Icon:

tone and volume

doorbell

psychoacoustics
Psychoacoustics is the study of human perception of sound, that is, the science of human physiological and psychological responses to sound (including speech and music).

hearing/voice range
hearing range

sonar heading and range

audio quantization
quantification process
audio quantization

basic concept
Sample Size – How many bits are used to store a sample. 16 bit common
Sampling rate : 8k, 16k, 32k, 44.1k, 48k sampling rate
Number of channels: mono, dual, multichannel
Bit rate calculation
Bit rate = sample rate × sample size × number of channels

What:

The sample rate is 44.1 kHz, the sample size is 16 bits, and the two channel PCM encoded WAV file

Bitrate = 44.1k × 16 × 2 = 1411.2kb/s = 176.4KB/s

audio compression
Audio compression is a type of data compression used to reduce the transmission bandwidth requirements of streaming audio media and the storage size of audio files.

compression method
lossless compression
All the information of the original file is preserved and there is no difference from the original file on playback.

Data compression through information redundancy is a reversible process.

lossy compression
Approximate some information in the original file to get a smaller file.

The incorporation of human psychology and recognition of the auditory system in the compression results is an irreversible process.

Audio signals outside the range of human hearing and audio signals that are masked.

masking effect
Masking effect: A phenomenon in which the auditory system’s perception of one sound is obstructed by another.

frequency domain masking
One sound is drowned out by another sound at the same time.

Mp3 Normalizer – Masking Effects Part 2

They are related to the frequency and relative loudness of sounds of different frequencies, whereas temporal masking is only related to time.

If two sounds are particularly close in time, humans may also have trouble telling them apart. For example, if a loud sound is followed by a very weak sound, the last sound will be difficult to hear. But if you play the second sound some time after the first sound stops, the last sound can be heard. How long should the interval be? For pure tones it is generally 5 ms. Of course, if the reverse effect is the same over time, if a lower sound appears before a higher sound, and the interval is too short, you won’t hear the lower sound.

Enter Bitrates, Stage Left La

JPEG compression can explicitly control the rate at which compression is discarded, but Mp3 users cannot. However, mp3 users can specify how many bits are used to store each second of music. The end effect is the same.

During encoding, the “garbage components” of the signal are compared to mathematical models of human psychoacoustics, as well as the bit rate used for compression, to decide which data to discard. The current bitrate used for mp3 compression is generally 128 kbps. The encoder will take this number into account when generating each data frame. If the bitrate is relatively low, the definition of “irrelevant” and “redundant” data will be relaxed, resulting in a large amount of data being considered useless data. Compressed audio will lose a lot of detail, resulting in a loss of sound quality. Conversely, if a higher bit rate encoding is used, the “irrelevance” and “redundancy” criteria are more precisely defined, details are preserved, but the file size is larger.

Please note that the bitrate of an mp3 file refers to the total bitrate of all encoded channels. That is, a 128 kbps stereo mp3 file is the same size as two 64 kbps mono mp3 files at the same time. But one 128kbps stereo file does sound better than two separate 64kbps mono files. Because in a stereo mp3 file, all the bits can be allocated (unevenly) to two channels as needed, for example, at a given time, one channel uses 60% of the bits and the other uses the remaining 40% , but the total number of bits will not exceed the bitrate parameter specified before encoding.

Fixed bit rate and variable bit rate

We assume that the mp3 encoding discussed here uses a fixed bitrate encoding method, which means that the output bitrate of the encoded file at any time period is whatever value you specify. The disadvantage of fixed bit rate encoding is that the amount of information in most sound files is not constant. In audio clips that use more musical instruments or have many people talking at the same time, the amount of information is large and vice versa, there are many factors that affect the amount of information in audio files. Variable bit rate encoding was developed to accommodate this characteristic of audio files. Variable bitrate encoding, which adjusts the bitrate used for encoding at any one time according to the dynamic characteristics of the audio data.

In most cases, variable bitrate encoding can achieve essentially the same sound quality as fixed bitrate encoding with a smaller file size. But variable bit rate encoding has its own drawbacks. First of all, some older players just don’t support decoding variable bitrate mp3 files and can’t play such files. Second, when the decoder plays a variable bitrate mp3, it cannot determine the current decode (play) position, and the “current play time” displayed on the player is inaccurate.

The information in the header of each frame is the same for a compressed fixed bitrate mp3 file, but not for the variable bitrate mp3 encoding. But when decoding, variable bitrate encoding does not require more computing power than a fixed bitrate file, because the mp3 decoder reads the full frame header even when playing a fixed bitrate mp3 file .

Mp3 Normalizer – Masking Effects

mp3 encoding: “masking effects” and bitrate

masking effects

The process of information filtering by human consciousness includes a process called “masking”. People who study psychoacoustics are very concerned with the study of this process, which studies the relationship between hearing, consciousness and sound. Mutual relationship. There are two masking effects that affect the Mp3 encoding process: one is acoustic masking and the other is temporal masking.

Sync masking (also called acoustic masking)

To describe the sync masking effect, it’s best to use an analogy. Imagine a bird flying in front of the sun. You see the bird fly from the left between you and the sun, and then the bird disappears because the sun’s rays are too bright. When the bird leaves the area of ​​the sun, you can see it again. Just like in a quiet environment, the sound of a guitarist’s fingers sliding across the strings can be heard, but if the same sound is played in an environment where rock music is being played, the average person cannot hear it.

The Mp3 codec only cares about the interrelationship between frequencies and loudness. Sync masking is described in terms that the mp3 codec can handle as follows: you have a sound signal, it’s a 1000hz sine wave (one), then we have a 1100hz sine wave (two), the wave sine two is weaker, -10db. Most people do not notice the presence of sine wave two in this situation. But sine wave two is not easy to perceive, not only because it is weaker, but also because its frequency is very close to that of sine wave one. To illustrate this phenomenon, we gradually increase the frequency of the second sine wave, but keep its volume constant until we can hear it. Assuming its frequency increases to 4000 Hz, we can hear this sound. When the frequency difference between the two sine waves gradually increases, the second sine wave can be heard gradually, until its frequency increases to a certain point, most people can hear two different tones, one is louder and the other is softer. lower.

This process is what psychoacousticists call the “synchronization masking” phenomenon. Two sounds with similar frequencies but very different volumes are difficult for humans to perceive as two different sounds. Taking this phenomenon into account, mp3 tries to discard those sounds that cannot be perceived during the encoding process, or assign as few bits as possible to these sounds.

Temporary masking Temporary masking Masking effects
synchronous