Examination of the Effect of Audio Signals on Neural Activity of Newborns

In this project we intend to examine the influence of music (of different kinds) on new-born babies.


In this project we intend to examine the influence of music (of different kinds) on new-born babies. We performed a frequency signal analysis using a spectrogram that presents the connection between the signal in time, its spectra and its intensity. The spectrogram is built using a fourier transform which gives a full spectra distribution at any given moment. In order to examine the influence of the music, we chose to check the level of compatibility of the spectral distribution vectors by correlating between them.

The problem:

One of the main problems in this project was cleaning the EEG signals and determining which are the relevent parts of it. Also we had to determine the way to approach the data and analyse the signal, so that we would be able to distinguish between musical parts and silent ones.

The basic approach:

We cleaned the spectrogram from disturbances using a thresh hold energy which was determined for each baby according to its average intensity level. In order to avoid any influence of the disturbances on the analysis, we considered the relevant vectors as zero.

In the first stage we chose two representative vectors (one from a quiet area, and one with music) and we checked their degree of correlation with the rest of the signal. We expected to find changes between the different time areas. We repeated this for all the vectors and took an average in order to achieve a more general understanding as to the nature of the correlation in the different areas.

There didn’t seem to be any clear changes, thus we decided to refine our analysis by checking, for each vector, the correlation with the rest of the vectors from its area only. This time we received much better results in which the influence is shown very clearly. The results are not entirely consistent- for some babies there was a considerable growth in the correlation level in the areas with the music (especially classical) and for others, there was a descent. But in most cases there was a noticeable change once the music began.

In the next stage we tried to focus on the transient phenomenon, which we identified between the different areas. This was done by correlating only the near surrounding for each vector by using a window of 20 seconds from each direction.


In this project we used the P.C stations in the VISL laboratory, and the following software: Matlab , Excel, Word and Nicolet.


We received encouraging results for most babies which show steep local maxima in the borders of the different areas and other results which imply a diffrence between the EEG signals in music and in silance. We believe that with a more accurate and consistant experiment it will be possible to achieve much more concrete results.


We are grateful to our project supervisor Hagai Kirshner for his help and guidance throughout this work, Johanan Erez our lab supervisor and to Doctor Zalman Weintraub for conducting the experiment and supplying us the data.