radio signals; philosophy of technology; digital corporeality; design research; posthuman subjectivity; machine learning; media studies
Radio Explorations Data
||Savic, Selena; Martins, Yann Patrick
|Persistent Identifier (PID)
The dataset is obtained from the Signal Identification Guide (SIGID) wiki, an organized database of information on radio signals. It contains audio samples (.wav, .mp3, .ogg), spectrogram images (.png) and metadata on radio signals in two .csv files. Each signal is characterized by signal type, frequency, bandwidth, modulation type, location, sample audio, spectrogram and a short description. They were received and recorded using software defined radio, and most of the audio samples have been demodulated from IQ energy information to audio. The dataset is created by crawling the SIGID wiki website to gather the data on known and unknown signals, updated last time in February 2021.We include two python scripts that can be used to produce audio chunks and corresponding spectrograms, which were used to train the self-organizing map models in the research projects.
This project proposes a humanities exploration of machine learning techniques, using methods characteristic for art and design research. Starting from an ambition to address the persistent and foundational dualism in modern thinking (e.g. nature - culture, subject - object, mind - body), the project engages with technology and computation in an experimental way. The experience of working with machine learning algorithms opens up a space of thinking which is characterized by synthesis in place of analysis: creating alloys rather than cutting apart. I propose to work with these processes as procedures for encoding and decoding information from noise (data), and demonstrate them as negentropic, synthesis-oriented ways to acquire knowledge. I will work with radio signals, both naturally occurring and man-made, and identify patterns, events, activities in the data on different types of transmissions. I will work with the data set from the Radio Signal Identification Guide Wiki. With this data, I will produce two case studies: the first one focusing on spectrogram images of all signals; the second one focusing on abstracting from existing categories in the dataset and reorganizing it according to subjective preferences of researchers. Browser-based tools or digital observatories for organizing and navigating this data in meaningful ways will be developed in scope of the project and used in both studies. These observatories will be presented at two workshops with invited experts. The design of digital observatories, workshops and their results will be documented in a peer-reviewed article. These outputs will address the articulation of similarities and difference beyond simple ontological distinctions onto matter or information, subjects or objects. I see this work as an experimental setup for further work on methods for engaging with abundant information. In the future, I hope to apply these methods on more complex datasets while continuing to explore digital corporeality.