The sound classes in the UrbanSound and UrbanSound8K datasets are taken from the urban sound taxonomy presented below. Click here to download a scalable PDF version of the taxonomy. For further information about the taxonomy please see our paper and the information provided below.
background & Design principles
The taxonomical categorization of environmental sounds, a common first step in sound classification, has been extensively studied in the context of perceptual soundscape research [1]. Specific efforts to describe urban sounds have often been limited to subsets of broader taxonomies of acoustic environments [2], and thus only partially address the needs of systematic urban sound analysis.
One of the first taxonomies for environmental sounds was proposed by Schafer [1], dividing sounds into six broad categories: natural, human, society, mechanical, silence and indicators. Gaver proposed a taxonomy based on different types of materials and the possible interactions between them [3]. Guastavino identified main categories of sound sources emerging from participants' descriptions of their ideal urban soundscape [4]. A similar study was conducted by Raimbault and Dubois
[5] who also studied the types of categorizations used by people to describe soundscapes. Brown built on previous proposals including [5], and proposed of the most detailed taxonomies for sounds in the urban acoustic environment [2]. For an exhaustive review of previous work the reader is referred to [6].
Given the various efforts to categorize environmental and more specifically urban sounds, it is not obvious which taxonomy we should work with, and whether it would satisfy the needs of our research. Indeed, based on the existing literature it seems that there isn't really any consensus as to how to build such a taxonomy, and different proposals employ different concepts to distinguish different classes in their taxonomy including physical sound reproduction, function and form.
In our view, an urban sound taxonomy should satisfy the following three requirements:
One of the first taxonomies for environmental sounds was proposed by Schafer [1], dividing sounds into six broad categories: natural, human, society, mechanical, silence and indicators. Gaver proposed a taxonomy based on different types of materials and the possible interactions between them [3]. Guastavino identified main categories of sound sources emerging from participants' descriptions of their ideal urban soundscape [4]. A similar study was conducted by Raimbault and Dubois
[5] who also studied the types of categorizations used by people to describe soundscapes. Brown built on previous proposals including [5], and proposed of the most detailed taxonomies for sounds in the urban acoustic environment [2]. For an exhaustive review of previous work the reader is referred to [6].
Given the various efforts to categorize environmental and more specifically urban sounds, it is not obvious which taxonomy we should work with, and whether it would satisfy the needs of our research. Indeed, based on the existing literature it seems that there isn't really any consensus as to how to build such a taxonomy, and different proposals employ different concepts to distinguish different classes in their taxonomy including physical sound reproduction, function and form.
In our view, an urban sound taxonomy should satisfy the following three requirements:
- It should factor in previous research and proposed taxonomies.
- It should aim to be as detailed as possible, going down to low-level sound sources such as “car horn” (versus “transportation”) and “jackhammer” (versus “construction”).
- It should, in its first iteration, focus on sounds that are of specific relevance to urban sound research, such as sounds that contribute to urban noise pollution.
Taxonomy construction
To address requirement (1), we decided to base our taxonomy on the subset of the taxonomy proposed by Brown [2] dedicated to the urban acoustic environment. We define 4 top level groups: human, nature, mechanical and music, which are common to most previously proposed taxonomies [1,2,4,6].
To address (2), we specify that the leaves of the taxonomy should be sufficiently low-level to be unambiguous – e.g. car “brakes”, “engine” or “horn”, instead of simply “car”.
Finally to address (3), we examined all the noise complaints filed through New York City’s 311 service from 2010 to date (over 370,000), and built the taxonomy around the most frequently complained about sound categories and sources: construction (e.g. jackhammer), traffic noise (car and truck horns, idling engines), loud music, air conditioners and dog barks to name a few.
To address (2), we specify that the leaves of the taxonomy should be sufficiently low-level to be unambiguous – e.g. car “brakes”, “engine” or “horn”, instead of simply “car”.
Finally to address (3), we examined all the noise complaints filed through New York City’s 311 service from 2010 to date (over 370,000), and built the taxonomy around the most frequently complained about sound categories and sources: construction (e.g. jackhammer), traffic noise (car and truck horns, idling engines), loud music, air conditioners and dog barks to name a few.
reading the taxonomy
Rounded rectangles represent high-level semantic classes (e.g. mechanical sounds). The leaves of the taxonomy (rectangles with sharp edges) correspond to classes of concrete sound sources (e.g. siren, footsteps). For conciseness, leaves can be shared by several high-level classes, indicated by a rectangle with an earmark.
future perspectives
Since the number of possible sound sources in an urban setting is very large (potentially infinite), we consider the taxonomy to be a constant work in progress rather than fixed. We plan to continue expanding and reformulating the taxonomy as we increase the scope of sounds covered by our research, by engaging the international research community and promoting a collaborative effort via (for instance) dedicated workshops.
references
[1] R. M. Schafer. The Soundscape: Our Sonic Environment and the Tuning of the World. Destiny Books, 1993.
[2] A. L. Brown, J. Kang, and T. Gjestland. Towards standardization in soundscape preference assessment. Applied Acoustics, 72(6):387–392, May 2011.
[3] W. W. Gaver. What in the world do we hear? an ecological approach to auditory event perception. Ecological Psychology, 5:1–29, 1993.
[4] C. Guastavino. The ideal urban soundscape: Investigating the sound quality of French cities. Acta Acustica United with Acustica, 92(6):945–951, 2006.
[5] M. Raimbault and D. Dubois. Urban soundscapes: Experiences and knowledge. Cities, 22(5):339–350, 2005.
[6] S. R. Payne, W. J. Davies, and M. D. Adams. Research into the practical and policy applications of soundscape concepts and techniques in urban areas. Department of Environment Food and Rural Affairs, HMSO, London, UK, 2009.
[2] A. L. Brown, J. Kang, and T. Gjestland. Towards standardization in soundscape preference assessment. Applied Acoustics, 72(6):387–392, May 2011.
[3] W. W. Gaver. What in the world do we hear? an ecological approach to auditory event perception. Ecological Psychology, 5:1–29, 1993.
[4] C. Guastavino. The ideal urban soundscape: Investigating the sound quality of French cities. Acta Acustica United with Acustica, 92(6):945–951, 2006.
[5] M. Raimbault and D. Dubois. Urban soundscapes: Experiences and knowledge. Cities, 22(5):339–350, 2005.
[6] S. R. Payne, W. J. Davies, and M. D. Adams. Research into the practical and policy applications of soundscape concepts and techniques in urban areas. Department of Environment Food and Rural Affairs, HMSO, London, UK, 2009.
