The comment of the paper: Skinput: Appropriating the Body as an Input Surface Skinput: Appropriating the Body as an Input Surface -- Rating: 5/5 This paper does a very great job about searching a novel human-computer interaction method. It perfectly jump out of the trap of the device size paradox: the trade-off between usability and mobility. It directs researchers who are in the dilemma to a new world. User’s skin, especially the arm skin focused of the paper is a very good interactive surface. It is available all the time without additional device. The tap input on the arm can be accurately captured by the sensor according to the experiment. So there is no doubt about the novelty and feasibility the of the idea. The methods used in implementing Skinput are quite fancy. As a very good example of cross-disciplines and put theories into practice, it involves bio-acoustics, signal processing, machine learning etc. Especially for the part of analyzing the waveforms and classify them to different inputs. 186 useful features are collected to give an accurate classification result. Different people may have some difference on their physiological structure of the arm. Training the classifier on-the-fly certainly fits the device to different users and sensor positions. The study in the paper did consider various different possibilities. They design three set of input locations, test with different positions of the sensors, test on both left and right arms. Different sets of input may be selected by future application designer to match with the feature of their product. Further, they even test the performance while users are walking and jogging. This is very meaningful since interaction certainly happens when users are moving. For example, we sometimes send a text or switch to another song while running. One problem I have with the paper is their failure to connect the function of the Pico-projector with the skinput. The projector has been mentioned in the article for several times, but I am still confused about the how the graphical interface is coupled with the input. All of the other part of the study are clearly present except this one. And the results are all around the accuracy. I think there should be more things to be analyzed than the classification accuracies, including the collaboration with the projector and the graphical interface.