Mayank Goel

People spend more than 90% of their time indoors, and, as such, improving indoor lighting quality can enhance their quality of life by positively influencing both physiological and psychological aspects. Numerous studies suggest that perceptions of a space vary depending on a number of lighting attributes present. Significant effort has been made across various fields to identify the spatial lighting conditions and attributes that impact human perception, although we lack studies that explore the impact of these conditions in tandem. This paper investigates how interior lighting conditions influence human impressions of room ambiance. The study examines 16 different interior scenes, considering factors such as sky condition (sunny vs. overcast), shading blinds position (open vs. closed), presence or absence of electric light (on vs. off), and Correlated Color Temperature (CCT) (2700 K, 4000 K, and 6500 K). The evaluation is conducted within an office environment at Carnegie Mellon University, using a combination of objective lighting metrics and subjective assessments. In total, 26 participants, consisting of 11 females, 14 males, and one undisclosed, aged between 18 and 50, evaluated the office ambiance under various lighting conditions using semantic differential scales. The analysis showed that the variation of blinds and CCT levels significantly influenced the participants’ impression of light. The study also identified statistically significant interactions between “blinds and CCT” and “blinds and sky” conditions, highlighting the combined influence of these variables on shaping indoor light impressions. This research offers valuable insights into the complex interplay of different lighting factors in shaping human perceptions, and underscores the importance of optimizing indoor lighting conditions for creating healthy and sustainable indoor environments.

Audio-based human activity recognition (HAR) is very popular because many human activities have unique sound signatures that can be detected using machine learning (ML) approaches. These audio-based ML HAR pipelines often use common featurization techniques, such as extracting various statistical and spectral features by converting time domain signals to the frequency domain (using an FFT) and using them to train ML models. Some of these approaches also claim privacy benefits by preventing the identification of human speech. However, recent deep learning-based automatic speech recognition (ASR) models pose new privacy challenges to these featurization techniques. In this paper, we systematically evaluate various featurization approaches for audio data, assessing their privacy risks through metrics like speech intelligibility (PER and WER) while considering the utility tradeoff in terms of ML-based activity recognition accuracy. Our findings reveal the susceptibility of these approaches to speech content recovery when exposed to recent ASR models, especially under re-tuning or retraining conditions. Notably, fine-tuned ASR models achieved an average Phoneme Error Rate (PER) of 39.99% and Word Error Rate (WER) of 44.43% in speech recognition for these approaches. To overcome these privacy concerns, we propose Kirigami, a lightweight machine learning-based audio speech filter that removes human speech segments reducing the efficacy of ASR models (70.48% PER and 101.40% WER) while also maintaining HAR accuracy (76.0% accuracy). We show that Kirigami can be implemented on common edge microcontrollers with limited computational capabilities and memory, providing a path to deployment on a variety of IoT devices. Finally, we conducted a real-world user study and showed the robustness of Kirigami on a laptop and an ARM Cortex-M4F microcontroller under three different background noises.