Research on key acoustic characteristics of soundscapes of the classical Chinese gardens

Wei Chen, Juanjuan LiuScientific 2023 · DOI

TL;DR

Researchers have decoded the "healing" frequency profiles of classical Chinese garden soundscapes, categorizing them into melodic "discrete" sounds (bird calls) and therapeutic "colored noise" (wind, rain, and water). By using LSTM neural networks and pitch detection, the study reveals how these specific acoustic signatures—rather than just volume—contribute to mental well-being and offer a blueprint for better urban green space design.

What Problem Does This Paper Address?

While we often think of gardens as visual masterpieces, classical Chinese gardens were designed as multisensory "sensescapes" where sound played a starring role. However, most modern soundscape research is obsessed with loudness (decibels). We know how to measure noise pollution, but we are less certain about what makes a "good" sound actually good.

The authors argue that frequency, pitch, and timbre are overlooked variables. They sought to understand the physical characteristics of the sounds preferred in historical Chinese literature and poetry—such as the "Orioles Singing in the Willow" or "Rain on Lotus Leaves"—and explain why these specific sounds have been prized for centuries as tools for healing and meditation.

Methods

The researchers combined historical humanities research with modern signal processing:

  1. Selection: They identified 12 representative soundscapes (4 biophonies/bird calls and 8 geophonies/wind, rain, water) based on extensive reviews of classical Chinese poetry and garden monographs like the Yuanye.
  2. Standardization: To isolate frequency from the confounding variable of loudness, they used high-quality samples from the BBC Sound Effects library rather than field recordings, which are often tainted by site-specific noise.
  3. Acoustic Analysis:
    • Pitch Detection: Used the VariAudio algorithm to map the frequency shifts in bird calls.
  4. LSTM Classification: They employed a Long Short-Term Memory (LSTM) neural network to classify continuous sounds. This is a sophisticated approach compared to standard spectral analysis, as LSTMs are better at identifying patterns in complex time-series data like "colored" noise.
  5. Software Stack: Matlab for extraction, Cubase 8 for noise reduction/pitch analysis, and Adobe Audition for spectral visualization.

Datasets

  • Historical Corpus: The study drew from the Book of Songs (Shijing), History of Chinese Classical Gardens, and several other regional historical texts to validate the "preference matrix" of sounds.
  • Audio Data: 12 categories of sound samples sourced from the BBC Sound Effects archive. While not a "new" dataset in the sense of a public benchmark, the processed 34-minute subset serves as the primary evidence for their frequency-based classification.

Key Insights and Ideas

  • The "Ear Before Eye" Philosophy: The paper highlights an ancient Chinese linguistic quirk: in idioms where "ear" and "eye" appear together, "ear" almost always comes first. This underscores a cultural priority for the auditory environment.
  • Discrete vs. Continuous: The researchers found a binary divide in preferred sounds.
  • Discrete sounds (bird calls) are essentially "proto-music." They exhibit two or more distinct pitch variations, creating a balance of surprise and predictability that triggers dopamine release.
  • Continuous sounds (wind/rain) act as "colored noise."
  • The Pink Noise Preference: Most of the preferred continuous sounds (6 out of 8) were classified as pink noise. Unlike white noise (which has equal energy across all frequencies and can sound "hissing"), pink noise has more energy at lower frequencies, which the human ear perceives as more balanced and "pleasant."

Key Results

  • Melodic Birds: All four bird species studied (Indian cuckoo, common cuckoo, red-crowned crane, and oriole) produced calls with specific musical intervals (e.g., the Indian cuckoo follows a pattern like $^#f^3-^♮f^3-^♮f^3-d^3$).
  • Noise Profiles:
    • White Noise: Small streams and wind in pine forests.
  • Pink Noise: Waterfalls, springs, wind in bamboo, and rain hitting various leaves (Musa, bamboo, lotus).
  • Healing Mechanism: The study concludes that these sounds provide "acoustic masking." By filling the frequency spectrum with soothing pink noise or engaging melodic pitches, the garden environment effectively "muffles" the chaotic, stressful sounds of the outside world.

Limitations and Future Work

The authors acknowledge a few key areas for growth:

  • Library vs. Field Data: While using BBC samples allowed for a "clean" frequency analysis, it lacks the spatial complexity (reverberation, distance) of a real garden. Future work should incorporate field recordings as a supplement.
  • Cultural Specificity: The study is rooted in Chinese aesthetics. The authors suggest that cross-cultural studies are needed to see if these frequency preferences are universal or culturally learned.
  • Timbre: The research focused heavily on pitch and frequency. The "texture" of sound—timbre—remains a complex frontier that requires more sophisticated modeling to fully understand why rain on a lotus leaf sounds "better" than rain on a tin roof.

Summary updated

Abstract

Soundscapes have played an important role in the design and building of classical Chinese gardens. In Chinese classical poetry, biophonies such as bird calls, and geophonies such as wind, are the preferable sound sources. Although these major sound sources have been categorized and summarized by scholars extensively, little research has been conducted to analyze the physical characteristics and preference matrix of these preferred sound sources. Moreover, the perceived loudness of sound in classical Chinese gardens has received more attention from scholars than acoustic frequency. In this study, we selected 12 sound sources that are most typically present in classical Chinese gardens based on extensive literature research on Chinese classical poetry, and acquired respective audio samples from the BBC's library of Sound Effects. Through the spectrogram analyses, pitch detection algorithm and LSTM audio classification methods, the sound sources were classified into discrete sound sources with pitch variation and continuous sound sources with spectral characteristics of white noise or pink noise. The reasoning behind the preference for these two types of sound sources was then discussed from physical and mental healing perspectives, which aims to help provide perspectives on the associated implications in the planning of urban green spaces.
BibTeX
@inproceedings{chen2023research,
  author = {Wei Chen and Juanjuan Liu},
  title = {{Research on key acoustic characteristics of soundscapes of the classical Chinese gardens}},
  booktitle = {Scientific},
  year = {2023},
  doi = {10.1038/s41598-023-39457-z},
}
Figure