Audio Processing is the use of computational, electronic or mathematical techniques to modify, analyse, generate, encode, transmit or otherwise manipulate audio signals. It can operate on recorded Digital Audio, live signals, speech, music, environmental sound, instrument signals or synthetic audio. Processing may occur in the time domain, frequency domain or another mathematical representation. Common operations include filtering, equalisation, dynamic-range compression, limiting, reverberation, delay, pitch shifting, time stretching, noise reduction, mixing and spatial processing. Audio Processing is used in music production, broadcasting, telecommunications, hearing technologies, games, film, speech recognition, scientific measurement and consumer electronics. The discipline combines signal processing, acoustics, psychoacoustics, mathematics, software engineering and electrical engineering.
Filtering is one of the most fundamental Audio Processing operations. A filter modifies the relative strength of different frequency components in a signal. Low-pass filters attenuate frequencies above a selected region, high-pass filters attenuate frequencies below it, band-pass filters preserve a band while reducing frequencies outside it, and notch filters strongly attenuate a narrow frequency region. Equalisation uses filters to alter tonal balance, while crossover networks divide frequency ranges between loudspeakers or processing paths. Filters can be designed with different frequency responses, phase behaviours, steepness and computational costs. Finite impulse response and infinite impulse response filters have different mathematical and implementation characteristics, and real-time systems must also consider latency and numerical stability.
Dynamic-range processing changes the relationship between signal level and output level. A compressor reduces gain above a chosen threshold according to a specified ratio, while an expander increases dynamic variation below a threshold. A limiter is essentially a high-ratio form of compression designed to prevent peaks exceeding a defined ceiling, although practical implementations vary. Attack and release times determine how quickly processing responds and recovers, which strongly affects the resulting sound.
A noise gate reduces or mutes signals below a threshold. These tools are widely used in recording, broadcast and sound reinforcement to control levels, improve intelligibility and manage headroom. Excessive processing can cause pumping, breathing, distortion or loss of natural dynamics, so technical correctness does not automatically imply an aesthetically desirable result.
Frequency-domain processing represents audio according to spectral components rather than directly according to time samples. The Fourier transform and related techniques allow systems to examine how energy is distributed across frequencies. Spectral processing can support noise reduction, equalisation, restoration, pitch analysis and sound synthesis. The short-time Fourier transform divides audio into overlapping windows to analyse how frequency content changes over time, producing a spectrogram-like representation.
Time-frequency resolution involves a trade-off because shorter analysis windows provide better temporal localisation but poorer frequency resolution, while longer windows do the opposite. More specialised transforms and adaptive methods are used for particular applications. These techniques are central to speech processing, music analysis, audio restoration and machine-learning systems.
Audio Processing is essential in communications systems. Speech codecs reduce the amount of data needed to transmit voice while attempting to preserve intelligibility and naturalness. Echo cancellation estimates and removes acoustic feedback caused when a loudspeaker's output is captured by a microphone. Noise suppression attempts to distinguish desired speech from environmental noise, while automatic gain control adjusts recording levels. Voice activity detection identifies periods containing speech and can reduce unnecessary processing or transmission. Modern conferencing systems combine these techniques with beamforming, dereverberation and source separation to improve communication in complex acoustic environments. Such systems must operate with very low latency because excessive processing delay makes natural conversation difficult.
Music and media production use Audio Processing creatively as well as technically. Reverb and delay simulate or construct acoustic environments, modulation effects alter timbre or spatial movement, distortion changes waveform characteristics, and pitch-shifting or time-stretching modifies musical material. Mixing combines multiple channels while controlling their levels, frequency relationships and spatial positions. Mastering applies final-stage processing intended to prepare a production for a particular distribution context. Digital audio workstations allow these processes to be automated, chained and recalled non-destructively. Plug-in standards allow third-party processors to operate within host applications, while virtual instruments generate audio from control data or software models. The distinction between corrective processing and creative processing is therefore useful but not absolute, because the same mathematical operation can serve either purpose.
Audio Processing developed from analogue filtering and recording techniques into increasingly sophisticated digital signal-processing systems as integrated circuits and computers became capable of real-time numerical computation. Digital recording enabled precise editing and complex effects without physically modifying tape, while fast processors allowed increasingly advanced software-based processing. Portable devices, smartphones and network communications subsequently embedded Audio Processing into everyday electronics. Machine learning has expanded capabilities in speech enhancement, source separation, transcription, synthesis and automatic mastering. Modern processors may execute billions of audio-related operations efficiently, but the basic principles of sampling, filtering, spectral representation and signal dynamics remain closely related to earlier signal-processing theory.
Audio Processing has limitations arising from noise, clipping, insufficient resolution, poor recordings, computational constraints and the information lost through irreversible transformations. A heavily clipped signal cannot generally be restored perfectly because the original waveform peaks have been replaced by distorted values. Aggressive noise reduction may remove legitimate signal components or produce artificial-sounding artefacts. Lossy codecs can discard information that later processing cannot recover. Processing can also introduce latency, phase changes or transient smearing depending on the algorithm.
Audio Processing is distinct from Digital Audio because Digital Audio refers primarily to digital sound representation, while Audio Processing concerns operations performed on sound signals. It is also related to audio engineering and sound design, but those broader disciplines include recording, acoustics, production and artistic decisions beyond signal manipulation itself.
