Clinical Depression Detection in Adolescent by Face
Prajakta Bhalchandra Kulkarni, Minakshee Patil · 2018 International Conference on Smart City and Emerging Technology (ICSCET) · 2018
Depression is a hidden and harmful disease in the world. World Health Organization (WHO) said that in 2017, 300 million plus humans are suffering from depression. The depression percentage in adolescents in India is in between 0.3 to 1.2. Hence it will cause a harmful impacts on their lives. Depression occurs due to many reasons like hectic daily routine, death of loved one, tension etc. Feeling of hopelessness, anxiety, no interest in any activity, loss or gain of weight are few symptoms of depression. If the stage of stress lasts more than one or two weeks then this stage is called as a depression. Depression may affect the social life, Health, suicidal thoughts, mentally disturbed if it takes a severe form. If the stage of depression is detected early then there will be a chance to save the person from a depression. When depression becomes more than the mild and moderate form, then that term is known as clinical depression. In our paper we introduces the method of depression detection which is harmless, easy and more accurate without the help of a psychiatrist. For implementation of a depression detection method, two algorithms were used named as Fisher vector algorithm and LTrP. Fisher vector is used for representation and description of an image. It uses Gaussian mixture model (GMM). Efficiency of fisher vector encoding is great for a computation. It gives a best result even with the linear classifier. We have applied this algorithms on a face. For the feature extraction LTrP is applied. Local tetra pattern uses a central pixel as a reference pixel with its neighbourhood pixel with respect to dimensions and then gives a magnitude as well as tetra pattern. for a better and accurate classification result fisher vector encoding is computed. Fisher vector encoding left the demerits of Bow that is bag of words. And LTrP left the demerits of Local binary pattern and gives the better results. This method gives the classification result in `Depressed' or `Not depressed' form.