Abstract A neural network algorithm has been applied to improve the energy spectrum and the energy resolution of a CdTe semiconductor detector. In this study, the three-layered perceptron has been chosen as a neural network. The input layer of the neural network consists of four neurons for amplitudes of pulses shaped by linear amplifiers with different time constants (0.5–6 μs), while the output layer consists of one unit for the correction factor by which the amplitude of a pulse shaped with a small time constant (0.5 μs) is multiplied to obtain the proper amplitude. The neural network has learned with the amplitudes of deficient pulses and the amplitude of an ideal pulse with no hole contributions and no effects of trapping and detrapping of the charge carriers. Then the network has been applied to correction of the amplitudes of shaped pulses of events with some hole contributions. The fairly good results obtained showed that the energy resolution and the photopeak efficiency can be improved by using the neural network algorithm.
[1]
P. Siffert,et al.
High resolution gamma ray spectroscopy with CdTe detector systems
,
1992
.
[2]
M. R. Squillante,et al.
Large, High Resolution CdTe Gamma Ray Sensors
,
1986,
IEEE Transactions on Nuclear Science.
[3]
Y. Eisen,et al.
Correction of incomplete charge collection in CdTe detectors
,
1994
.
[4]
E. Caroli,et al.
Computer simulation of charge trapping and ballistic deficit effects on gamma-ray spectra from CdTe semiconductor detectors
,
1994
.
[5]
M. Martini,et al.
Trapping and detrapping effects in lithium-drifted germanium and silicon detectors
,
1970
.