How iOS 15 Filters Background Noise on Calls

Apple introduced Voice Isolation in iOS 15 to significantly improve call clarity by stripping away distracting ambient sounds like lawnmowers, traffic, and barking dogs. This feature relies on advanced machine learning models powered by Apple’s Neural Engine, alongside multi-microphone hardware processing, to distinguish human vocal patterns from environmental noise in real time. By prioritizing vocal frequencies and suppressing non-human sounds, iOS 15 ensures that listeners hear only the speaker's voice, even in chaotic environments.

The Role of On-Device Machine Learning

The core of iOS 15’s noise suppression is a deep neural network trained on thousands of hours of audio data containing both human speech and common ambient noises. The model learns the acoustic characteristics of the human vocal tract, such as pitch, resonance, and harmonic structures.

When audio enters the microphone, Apple's A-series Bionic chip uses its Neural Engine to analyze the audio waveform in milliseconds. Because the processing occurs entirely on the device, there is virtually no latency, and personal audio data remains private.

Identifying and Isolating Sound Profiles

Ambient sounds have distinct spectral signatures that differ sharply from human speech:

  • Lawnmowers and Engines: These produce continuous, low-frequency acoustic energy with repetitive mechanical patterns. The machine learning model identifies these consistent waveforms and applies dynamic filtering to cut the specific frequency bands where the motor operates.
  • Barking Dogs and Sharp Disruptions: Unlike continuous hums, barking dogs represent transient, high-amplitude bursts with irregular intervals. Traditional noise gates often fail to catch these sudden spikes. The neural network in iOS 15 recognizes the sharp attack, decay, and frequency profile specific to canine barks, instantly attenuating those specific sonic peaks without muting the caller's ongoing speech.

Hardware Beamforming and Spatial Filtering

Software algorithms work in tandem with the iPhone's multi-microphone array. iPhones use beamforming technology—combining input from microphones located at the bottom, front, and back of the device—to create a directional audio field focused directly on the speaker's mouth.

The microphones capture both the target voice and the ambient environment from different physical angles. The system calculates the phase and arrival time differences of the incoming sound waves. Sounds originating outside the direct path of the speaker's mouth are tagged as environmental noise, allowing the software to subtract them from the primary audio feed before the signal is transmitted.

How to Enable Voice Isolation

Voice Isolation does not run by default on standard cellular calls in iOS 15 (it was initially limited to FaceTime and compatible VoIP apps like WhatsApp and Zoom). To enable it during a call:

  1. Swipe down from the top-right corner of the screen to open the Control Center.
  2. Tap the Mic Mode tile in the upper-right corner.
  3. Select Voice Isolation from the available options (Standard, Voice Isolation, and Wide Spectrum).

Once activated, the system dynamically filters out background disturbances, keeping vocal output clear and intelligible.