Homeglossary

Noise Suppression

Noise suppression is an audio signal-processing technique that reduces unwanted background sound while preserving useful speech or other target audio. It is used in microphones, calls, recordings, hearing devices, and communications systems to make the intended sound easier to hear. Its results depend on the noise type, microphone placement, processing method, and how well the system separates speech from similar sounds.

What Is Noise Suppression?

Noise suppression is audio processing that reduces unwanted background sound while keeping the main sound, usually speech, as clear as possible. It is used in microphones, calls, recordings, hearing devices, and voice-controlled systems.

The target may be speech, music, or an alert tone. Noise can include fans, traffic, typing, electrical hum, wind, room echo, or nearby voices. Suppression reduces rather than completely removes noise because real-world sounds can overlap with speech in frequency and timing. The goal is clearer audio without an unnatural voice.

A steady air conditioner is relatively predictable and often easier to lower. A nearby speaker is harder to remove because that voice resembles the intended speaker.

How Noise Suppression Works

Noise suppression can run in real time during a call or after recording. Traditional digital signal processing uses filters and audio rules, while newer systems may use AI models trained to recognize speech patterns.

  1. A microphone captures the intended voice and background sounds together.
  2. The system examines frequency, volume, timing, and audio changes to identify likely noise.
  3. It estimates a noise profile, such as fan noise, hiss, electrical hum, or keyboard clicks.
  4. The processor reduces components likely to be noise while trying to preserve speech.
  5. The cleaned signal is sent to a call, recording, stream, or hearing device.
  6. AI-based speech isolation can estimate whether sound resembles speech, but may make mistakes when voices overlap.

Common Types of Background Noise

Predictable sounds are usually easier to reduce than sudden sounds or sounds that resemble the target voice.

Noise typeExampleTypical suppression difficulty
Steady noiseFan, air conditioner, computer humUsually easier because the pattern changes slowly.
Intermittent noiseDoor closing, dishes, notificationsModerate to difficult because sounds start suddenly.
Wind noiseOutdoor recording or air hitting a microphoneOften difficult without physical wind protection.
Room reverberationEcho from bare walls and hard floorsDifficult because it is a delayed copy of the speaker's voice.
Nearby speechConversation in the same roomVery difficult because it has speech-like characteristics.
Keyboard soundsTyping during a video meetingOften manageable, but heavy typing can affect speech clarity.
Electrical humPower interference at a low toneOften easier when the hum is stable and identifiable.

Noise Suppression vs. Noise Cancellation, Noise Reduction, and Noise Gates

These related terms describe different tools and goals. Noise suppression is not the same as active noise cancellation in headphones.

MethodMain purposeHow it works
Noise suppressionImprove microphone audioAnalyzes an incoming signal and lowers likely background noise while speech continues.
Active noise cancellationReduce what a listener hearsHeadphones use microphones and opposing sound waves to reduce external sound at the ears.
Noise reductionClean recorded audioEditing software applies filters or a sampled noise profile after recording.
Noise gateMute low-level sound during silenceCloses or lowers the channel when audio falls below a chosen volume threshold.

A gate does not distinguish typing from a voice while someone speaks. It mainly helps between words or sentences. Noise suppression is more selective, though quality depends on the software and audio conditions.

Key Components That Affect Audio Quality

Good results begin before software processing. Microphone directionality matters because a cardioid microphone favors sound from the front and rejects some sound from behind. Speaking close to the microphone raises voice level relative to room noise, allowing lower gain and less aggressive processing.

Hard surfaces reflect speech and create reverberation that software may confuse with noise. Sampling quality, processing delay, noise profiles, and voice-detection models affect clarity and latency. Better capture reduces the need for cleanup, a useful principle when planning app design and development for voice features or spoken-input tools.

Where Noise Suppression Is Used

Noise suppression supports communication wherever microphones operate outside quiet studios.

  • Video meetings and phone calls, to reduce household, office, and street noise.
  • Live streaming and gaming, where audiences need to hear commentary over keyboards and fans.
  • Podcasts, interviews, and journalism when recording cannot be repeated in a quieter place.
  • Call centers and customer support systems, where clearer speech can reduce misunderstandings.
  • Accessibility tools and hearing devices that may prioritize speech in noise.
  • Voice assistants and voice-enabled applications, including products built with an AI app builder, where reliable speech capture affects user experience.

Real-time suppression must avoid noticeable delay. Recorded audio can receive slower, more detailed cleanup during editing.

Benefits of Noise Suppression

When used appropriately, suppression improves audio usefulness rather than simply making it quieter.

  • Clearer speech can make calls, meetings, and recordings easier to follow.
  • Fewer distractions help listeners focus on the intended message.
  • People may need fewer repetitions of important details.
  • Usable recordings can be made in less-than-quiet locations.
  • Background sounds may be less audible, offering some privacy protection.
  • Creators can spend less time repairing consistent noise.

Limits and Common Audio Artifacts

Noise suppression is an estimate, not perfect separation. Strong settings can improve quiet moments while damaging speech during busy ones.

  • Clipped syllables can occur when word beginnings or endings are mistaken for noise.
  • Metallic, watery, or robotic voice quality can follow aggressive processing.
  • Pumping occurs when background sound rises and falls as the system reacts.
  • Processing can introduce latency that makes live conversation less natural.
  • Music, laughter, overlapping speech, and television audio can be misclassified or partly removed.
  • Another voice may remain audible because it overlaps the intended speaker's frequency range.

No tool can always identify the important voice without extra context, separate microphones, or physical separation between speakers.

How to Reduce Microphone Background Noise Before Processing

Start with the physical setup. Software works best when it has less unwanted sound to fix.

  1. Move closer to the microphone while avoiding breath bursts and distortion.
  2. Lower microphone gain to capture less room sound.
  3. Choose a suitable pattern, such as a front-focused pattern for one speaker.
  4. Move fans, vibrating devices, and other noise sources farther away when possible.
  5. Use curtains, rugs, furniture, or acoustic treatment to reduce echo.
  6. Use a foam cover or windscreen outdoors or when moving air reaches the microphone.
  7. Wear headphones during calls to prevent speaker audio feeding back into the microphone.
  8. Test with realistic sounds, including typing and conversation, before relying on a setting.

Choosing When to Turn Noise Suppression On or Off

Turn suppression on when steady room noise, typing, or distant activity makes speech harder to understand. Start moderately and increase it only if listeners still struggle. Maximum suppression is not always best because it can remove quiet words and make voices sound processed.

Consider turning it off or using a light setting for music, singing, instrument recording, or any session where natural tone matters more than cleanup. It may also be a poor fit when several people share one microphone, since the system may favor or alter voices.

A Practical Test for Evaluating Noise Suppression

Record short before-and-after samples in the actual environment. Include normal speech, pauses, typing, a fan or other steady noise, and a second person speaking at a distance. Compare speech naturalness, word intelligibility, missing syllables, residual noise, and delay.

The best setting is usually the lowest level that makes the target voice consistently understandable. If speech sounds less trustworthy or harder to follow, improve the microphone setup before applying more processing.

Start Building
on Emergent today
Start Building