Adjusting to the noise environment.
A team of European researchers has developed a new instrument to
measure and identify noises at work and in our daily lives. Using
improved computer technology, the PC based system compares the noise
against a database of 2,000 sounds. The prototype has been thoroughly
tested in the laboratory and in the field. It will soon be sold
as a commercial product and will make noise engineers' jobs easier.
What does pollution mean to you? Perhaps the first thing you think of is the smoke bellowing out of a factory chimney or an oil spill. Not many people would mention the loud drilling in the street or the noisy machines in factories. But noise pollution at work and at home can cause hearing problems, not to mention making your life a misery.
Noise control engineers are called out if someone makes a complaint about noise. They also perform routine sound checks. It can be difficult to find the source of the noise, especially because the acoustic level fluctuates and there are often many different sounds combined together. To do their job properly, the engineers need to build up a detailed picture of the noises - or as they call it, the acoustic environment - to which people are exposed.
"One of the most troublesome problems is that highly trained staff are needed on site to identify the sources of noise," says Dr Lefevre of Agora Conseil in France. "Obviously, the noise could be tape recorded but then it would have to be analysed back in the laboratory. The ultimate in noise measurement is to have an automatic system which can identify the type of noise and classify it into the various categories where action can be taken." If there was such an instrument, noise engineers would be able control noise annoyance more easily, both in daily life and in the workplace.
The MADRAS - Methods for Automatic Detection and Recognition of Acoustic Sources - project has made such a system a reality. In a two-year study, the project partners from across Europe developed a new generation of instrument that can automatically identify and measure, in real time, the different sound sources that make up the acoustic environment and measure their impact.
makes it possible
The effects of noise on human beings are extremely complex. They are related to objective criteria such as duration, repeatability, variation over time and frequency, and also to subjective criteria according to psychological, sociological and physiological contexts.
Developments in measuring environmental noise rely heavily upon advances in computing capabilities. About 15 years ago, researchers took advantage of improved technology to develop a method known as "short Leq" which allows sounds to be analysed by their acoustical signature. This approach is now used worldwide and has revolutionised noise measurement. However, it still has its shortcomings as the parameters it measures do not give a complete picture of the annoyance.
More recently, computing technology, data storage and microelectronics have provided the means to develop more accurate methods. Huge amounts of data on simultaneous observations can now be computed and analysed in real-time and this means that the effects of a given noise source can be quantified even when many noises are present. In addition, subsequent high-level processing involving artificial intelligence allows researchers to start to think about automatically identifying different sources by efficient comparison with sound databases.
The MADRAS project, coordinated by Dr Lefevre, decided to seize the opportunity that improved technology had opened to it. Investigating and adapting the technology and then developing and evaluating the prototype required a broad range of skills. "To achieve the objective of the project, the MADRAS consortium had to exhibit complementary expertise in acoustics, signal processing, artificial intelligence, statistics, system development, field testing and
marketing," says Dr Lefevre.
A key player in acoustics in Europe, 01dB in France, played a major role in creating the hardware and software for the prototype instruments. The SME, SINUS from the eastern part of Germany, assisted by developing specific hardware for data acquisition and the signal processing core engine. GRAS Sound and Vibration in Denmark handled the transducer studies, while statistical assessments were completed by CSTB, a public research laboratory in France. The Laboratory of Building Science at ENTPE in France worked on the artificial intelligence aspects of the system.
How does the system work? During the period of analysis, properties of the sounds that the system "hears" are compared to an extensive database of over 2,000 different sounds. This relies upon automatic source recognition algorithms and a sound database available on three CD-ROMs. CEDIA, a Belgian industrial laboratory linked to the Université de Liège, compiled this database.
A different class of noise
First, the sound is classified in quite broad terms. It could, for example, be classed as being a stationary source, impulsive noise, speech and shouts, isolated vehicles or heavy carriers. To do this effectively, the computer programme asks the engineer simple questions about the environment. The nature of each source is then identified in as much detail as possible using approaches that are specially suited to each of the different classes. Identifying the sound source is even possible if there are other sounds corrupting the noise in question.
Tested in the field by a potential user, Estudi Acustic - a Spanish SME with acoustic consultants - the final product can be used either for on-site measurements or for long-term monitoring. It is able to correctly identify the vast majority of sources that make up typical noise pollution in factories, in offices, at home and on the street.
While some of the instrument's modules still require further work, the system is nearly ready to be launched commercially. The first beta instruments in the series, which are a direct result of the project, should be available in the market place in the coming months.