Handwriting recognition has become one of the forthcoming research topics due to the exponential growth of the usage of the devices such as take note, PDA, iPad, and smart phonesetc. Writing on those devices is quite similar to the writing with pen and paper. But, these devices can take input faster than a keyword and also minimize the chances of errors due to mistyping. In this study, a model for Amharic handwriting character recognition is proposed for the 238 basic Amharic characterswritten by different background and age group. It is collected by using an Acer desktop with interactive board, LG tablet and Digimemo devices in UNIPEN format.It consists of 36220 character samples with 181 writers. Data need to be pre-processed, as the data collected in real environment, it can be noisy and inconsistent. It consists normalization, interpolation, smoothing and resampling techniques used and informative and non-redundant features builds by hybrid techniques like:-curliness linearity, slope , curvature, zones and aspect andonce this stage is provided with the data set from the feature extraction stage, those data can be fed to the decision making model to perform the recognition and classification. Our paper presents a model for Amharic handwriting character recognition by deep fully connected network (Deep-FCN).Extensive experimental results show that our model not only generates recognition of characters but also facilitates the accuracy improvement of handwritten character recognition rateon test dataset; the technique has been provided 96.56% recognition accuracy. The result thus obtained in this study is promising to apply fully connected network for character recognition. It is a further research direction. However, time constraint, only basic characters and character level recognition considered. But their solutions have not been tested in real environment. Keywords: Handwritten character recognition,Amharic handwriting character recognition, Stroke,deep fully connected network, OCR, CNN, SVM.