Traditional space defined as emotion container are places such as archives, museums, cathedrals, palaces, cemeteries, and memorials. However, Gare de Lyon, as an urban public space, is undoubtedly a social infrastructure that witnesses goodbyes and hellos, and embeds various emotions every second. On the other hand, considering the high density and fast flow of people in the station, as well as the crowd, noisy and confused physical environment, it is not surprised that among all kinds of the emotions, a large number of population are showing passive moods during transit. In this way, how to provide a joyful built environment that is able to comfort people with negative emotions becomes the design problem of the project Ungrumpy Paris.
The project will focus on identifying people with negative emotions in the station, and propose a design approach that could engage the emotion potential of the space as well as create a more human-oriented built environment.
The project consists of two parts. The mapping/surveying part uses images uploaded on social media as the major data resource to map the emotion of people. By image recognition, the computation method will be able to geo-locate the places where people are most likely to show interest on and take pictures. By doing this, places that have design potentials can be mapped out, and be used as a design foundation for further development (in this project, the area in front of timetable and the platform).
The design part proposes an interface – Ungrumpy Paris – as an intervention to provide comfort, joy and happiness for people. Located above the timetable and embedded in the electronic gate, the Ungrumpy Paris can detect people’s smile on face, and return a stylized portrait with the smile highlighted on it. Meanwhile, a sentence praising the smile will be returned for the user, in order to encourage more people to express their positive feeling, and pass this emotion to other commuters. The data containing the proportion of real-time smiling people is also able to be collected by the device, and be used to tell the emotion situation of the whole station at different time scales. The data can also be compared with other data sources, indicating what property (such as weather, the delay situation of the train, etc.) may influence people’s emotion, and allows doing comparisons across stations.
4. Where Ungrumpy Paris is located in the city
The project uses smile to identify happy people. Using facial recognition, the Ungrumpy Paris can detect the positive expression on people’s faces when waiting for the train under the timetable or boarding the train via electronic gates. Once the smiling faces are recognized, it will allow SNCF to have an idea of happy people’s proportion in the station. Simultaneously, the smiles of people who wear happy faces will be returned and highlighted on the screen, along with a sentence saying “T’es beau!/ Tu es belle“, as an encouraging mechanism praising people’s happiness. A key point is that rather than trying to get people’s consent of using and displaying their facial information in front of public, the system will firstly convert the real-time facial information into a stylized image. In this way, it can avoid the troubles caused by privacy. Hopefully, with the believe that the smile can spread, once a grumpy commuter witnesses his/her smiling neighbor being praised by the system, he/she will ideally have the same desire to play with the interface, with a propensity of showing a happy face.
Since the similar interface will be both embedded above the timetable panel and within the electronic gate, people will have the opportunity to interact with these devices twice. In this way, the emotional data for same person can also be collected twice, and to be used to analyze whether the expression of him/her has been changed after interacted with the Ungrumpy Paris, indicating whether the device is useful in ameliorating people’s negative emotions to some extent. By doing this, the project is able to close the feedback loop. The practical value of the design is to ameliorate people’s bad temper when weather is bad, luggage is heavy, station is crowded or the train is delayed, improve the emotional condition within the train station, meanwhile to solve potential conflict between the travelers and the administrators. On the other hand, the workers, cleaners and shop owners can also have the opportunities to play with the interface. This helps create a more comfortable and relaxed public environment.
The device of Ungrumpy Paris will be implemented in different train stations owned by SNCF, spreading all over the great Paris area. In this way, they make up a platform.
Such emotion detection platform can generate a large amount of data. For each station, the front end will show the emotion distribution of the station generally, e.g., the real-time proportion of happy people within station, etc.
Meanwhile, the emotion data profile can be compared with other datasets, e.g., the weather data, the noise data, the train delay data, etc. This will allow SNCF to run the analysis models (e.g., linear regression model) to examine which feature(s) will influence people’s emotion within the station mostly. Such information analysis and representation will take place at an urban scale, and will add another layer to the existing information of the station. After analysis, SNCF will be able to propose different methods in order to improve user experience within station based on their analysis result. For instance, if the research finds that bad weather, e.g., rain, mostly affects people’s emotion, SNCF can conduct more artificial light and change the humidity within station accordingly when weather is turning bad.
The technology that is embedded within this project is mainly based on current facial and expression recognition techniques.
Firstly, the images containing expressions which are already categorized are input as the expression dataset. The Haar Cascade classifiers are then imported, in order to set up the classification standard for the expression recognition. Afterwards, the system will convert the image(s) which is directly loaded or captured by external camera(s) to gray scale image(s). Then the internal facial detector established by the classifier will be used for the facial recognition, detecting every pixel of the image(s), comparing the relationship between the pixel itself and its neighbors with classifier. Thus, the face can ideally be recognized.
When finishing the first step, the system begins to categorize the expression, based on the CNN model which is built following the same logic of facial recognition. The dataset containing the classified expressions, including angry, fear, happy, sad, surprise, and neutral will be implemented as classification standard. Using the similar comparison methodology, the information of expression displayed on the image(s) can be categorized and returned.
Based on certain machine learning algorithm, the Ungrumpy Paris will take advantages of the facial and expression recognition technology in the station. For further development, similar technology can also be used to examine whether the traveler is coherent with the ticket owner or not, whose information will be previously collected by the ticket selling system, in order to improve the security of the station.
To investigate people’s emotions within station, especially within a station in France is indeed a great challenge, since French is famous with their grumpy expressions. In the same time, a large proportion of French are unwilling to share their emotion with public. However, with the scope of this project, the Ungrumpy Paris provides a potential alternative of sharing personal feelings with others, and getting a positive feedback by such interactions.