The last words of this short introduction to sound analysis concern sound amplitude measurement. As stated above, the amplitude for a given frequency is represented in spectrograms by a grayscale value. Each grayer step on this scale represents an increase in amplitude, which is generally expressed in db (decibels). The decibel scale can be used to express three different physical quantities: sound power (the energy radiated by a sound source in the unit of time), sound intensity (the energy passing through a surface perpendicular to the direction of sound, that is proportional to sound power), sound pressure (the small alternating change in ambient pressure caused by sound, that is proportional to sound intensity). Our microphone measures the sound pressure that is translated into a continuous varying electrical signal, whose amplitude is proportional to sound pressure. Decibels are the logarithm (base 10) of the ratio between the measured sound pressure value and a reference value (20 microPascal). This reference value correspond to the threshold of audibility of a 1000 Hz tone for a human (Bibliography 4). When recording bird songs it is obvious that sound pressure values may be heavily influenced by many factors: distance between the microphone and the sound source, presence or absence of obstacles, wind speed and direction, the direction of emitted sound etc. Therefore the ABSOLUTE values of sound amplitude measured by spectrograms aren't so meaningful. More interesting is the comparison between RELATIVE amplitudes associated with different frequencies into the SAME spectrogram. This relative pattern is often characteristic for a given song.
Fig. 8 shows two phrases extracted by the same song of a Cirl Bunting (Emberiza cirlus). In both spectrograms the maximal amplitude values (darker zone of the spectrogram, inside the two blue lines) are found between 4000 and 5400 Hz. The red dots indicates the dominant frequencies (= frequencies whose amplitude is maximal at that particular moment). It is evident that the frequency band associated with maximal amplitude doesn't change in these two song phrases of the same bird.
We can conclude that spectrographic analysis is a very useful tool for studying characteristics and biological significance of bird songs. Adding the correspondent waveform signal for comparison improves the efficacy of this technique. The main parameters of spectrograms (sampling frequency, FFT size, windowing type etc.) should be always shown to allow a better understanding of the results obtained. To get a deeper knowledge of spectrograms and digital sound analysis, I recommend the papers listed in Bibliography (2, 3, 4, 5).
Previous page - return to top