Signal Processing Concepts and Engineering Insights. 


Explore signal processing concepts, algorithm comparisons, and practical engineering insights.
Topics include FFT vs STFT, FRF analysis, filtering techniques, and other signal processing methods used in real engineering workflows. 

Advanced Analysis & ApplicationsWhy Human Hearing Is Logarithmic: From Hz to Perception

Why Human Hearing Is Logarithmic: From Hz to Perception

Human hearing exhibits a nonlinear response to both frequency and amplitude. While physical signals are measured on linear scales, our perception operates approximately on logarithmic scales.

Illustration showing logarithmic human hearing perception with a comparison between Hz and Mel scale and a human ear diagram

This distinction is critical in engineering fields such as

  • Audio signal processing
  • Acoustic measurement (NVH)
  • Speech recognition
  • Audio compression

Understanding how humans perceive sound allows engineers to design systems that align with perception rather than raw physical quantities.


Frequency Perception: Linear in Physics, Logarithmic in Perception

Although frequency is measured in Hertz (Hz), human perception of pitch follows a logarithmic pattern.

This means

  • Equal ratios in frequency are perceived as equal pitch intervals
  • Not equal differences

For example

  • 100 Hz → 200 Hz (×2)
  • 1000 Hz → 2000 Hz (×2)

Both are perceived as similar pitch steps, even though the absolute difference is very different.


Why This Happens

The human auditory system, particularly the cochlea, processes frequency in a nonlinear way.

  • Lower frequencies are resolved with higher precision
  • Higher frequencies are compressed perceptually

This leads to perception being based on frequency ratios rather than absolute differences


Perceptual Frequency Scales

To model human perception, several perceptual scales are used

BandDescriptionApplication
Octave bandBased on doubling of frequencyMusic, acoustics
Mel bandPerceptually uniform pitch scaleSpeech processing
Bark bandBased on critical bands of hearingPsychoacoustics


Mathematical Insight

1-Octave scale: based on double of frequency
1-Octave scale


Amplitude Perception: Loudness is Logarithmic

Just like frequency, amplitude perception is also logarithmic.

Loudness is approximately proportional to

Loudness

This is why sound levels are measured in decibels (dB), not linear amplitude. 


Practical Meaning

  • Doubling amplitude does not double perceived loudness
  • A 10× increase in power corresponds to a fixed increase in dB

This makes logarithmic scaling far more suitable for representing human hearing.


Practical Implications in Engineering

This logarithmic perception directly affects system design.

Audio Compression (MP3, AAC)

  • High-frequency components can be compressed more aggressively
  • Based on reduced perceptual sensitivity


Speech Processing

  • Mel-Frequency Cepstral Coefficients (MFCC) rely on Mel scaling


Equalization and Mixing

  • EQ bands are spaced logarithmically
  • Matching perceptual resolution


NVH and Acoustic Analysis

  • Human perception must be considered alongside physical measurements


MALMIJAL Example: Octave Band Analysis

Linear vs Logarithmic Scale in 1-Octave Band

linear vs. log-scale frequency of 1-octave band


PSD (Linear frequency scale)PSD (Power Spectral Density) of white noise in dB/Hz

PSD (Power Spectral Density) of white noise in dB/Hz


1/3-octave Band (Logarithmic frequency scale)

A-weighted 1/3-octave band level in dBAA-weighted 1/3-octave band level in dBA


1-octave Band (Logarithmic frequency scale)

A-weighted 1-octave band level in dBAA-weighted 1-octave band level in dBA


Comparison of 1-Octave and 1/3-Octave Band Analysis

1-octave level is higher than 1/3-octave level

1-octave band level is higher than 1/3-octave


Key Insight

Engineering systems that interact with human perception must not rely solely on linear physical measurements.

Instead, they should incorporate perceptual models such as

  • Logarithmic scaling
  • Mel or Bark transformations
  • Decibel representation

Bridging this gap is essential for creating systems that sound correct, not just measure correctly.


Suggested Further Reading

You may also find these topics helpful: