More Like this

1. Kirigami: Lightweight Speech Filtering for Privacy-Preserving Activity Recognition using Audio

   https://doi.org/10.1145/3643502  

   Boovaraghavan, Sudershan; Zhou, Haozhe; Goel, Mayank; Agarwal, Yuvraj (March 2024, Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies)

   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. 

   more » « less
2. Test-Time Adaptation Toward Personalized Speech Enhancement: Zero-Shot Learning with Knowledge Distillation

   https://doi.org/10.1109/WASPAA52581.2021.9632771  

   Kim, Sunwoo; Kim, Minje (October 2021, 2021 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics (WASPAA))

   In realistic speech enhancement settings for end-user devices, we often encounter only a few speakers and noise types that tend to reoccur in the specific acoustic environment. We propose a novel personalized speech enhancement method to adapt a compact denoising model to the test-time specificity. Our goal in this test-time adaptation is to utilize no clean speech target of the test speaker, thus fulfilling the requirement for zero-shot learning. To complement the lack of clean speech, we employ the knowledge distillation framework: we distill the more advanced denoising results from an overly large teacher model, and use them as the pseudo target to train the small student model. This zero-shot learning procedure circumvents the process of collecting users' clean speech, a process that users are reluctant to comply due to privacy concerns and technical difficulty of recording clean voice. Experiments on various test-time conditions show that the proposed personalization method can significantly improve the compact models' performance during the test time. Furthermore, since the personalized models outperform larger non-personalized baseline models, we claim that personalization achieves model compression with no loss of denoising performance. As expected, the student models underperform the state-of-the-art teacher models. 

   more » « less
3. A deep dive into microphone hardware for recording collaborative group work

   https://doi.org/10.5281/ZENODO.6852954  

   Bradford, Mariah; Hansen, Paige; Ross, J. Beveridge; Krishnaswamy, Nikhil; Blanchard, Nathaniel (January 2022, Educational Data Mining Conference)

   Mitrovic, Antonija; Bosch, Nigel (Ed.)

   Classroom environments are challenging for artificially intelligent agents primarily because classroom noise dilutes the interpretability and usefulness of gathered data. This problem is exacerbated when groups of students participate in collaborative problem solving (CPS). Here, we examine how well six popular microphones capture audio from individual groups. A primary usage of audio data is automatic speech recognition (ASR), therefore we evaluate our recordings by examining the accuracy of downstream ASR using the Google Cloud Platform. We simultaneously captured the audio of all microphones for 11 unique groups of three participants first reading a prepared script, and then participating in a collaborative problem solving exercise. We vary participants, noise conditions, and speech contexts. Transcribed speech was evaluated using word error rate (WER). We find that scripted speech is transcribed with a surprisingly high degree of accuracy across groups (average WER = 0.114, SD = 0.044). However, the CPS task was much more difficult (average WER = 0.570, SD = 0.143). We found most microphones were robust to background noise below a certain threshold, but the AT-Cardioid and ProCon microphones were more robust to higher noise levels. Finally, an analysis of errors revealed that most errors were due to the ASR missing words/phrases, rather than mistranscribing them. We conclude with recommendations based on our observations. 

   more » « less
4. CommanderGabble: A Universal Attack Against ASR Systems Leveraging Fast Speech

   https://doi.org/10.1145/3485832.3485892  

   Zhang, Zhaohe; Yang, Edwin; Fang, Song (December 2021, Annual Computer Security Applications Conference (ACSAC))

   Automatic Speech Recognition (ASR) systems are widely used in various online transcription services and personal digital assistants. Emerging lines of research have demonstrated that ASR systems are vulnerable to hidden voice commands, i.e., audio that can be recognized by ASRs but not by humans. Such attacks, however, often either highly depend on white-box knowledge of a specific machine learning model or require special hardware to construct the adversarial audio. This paper proposes a new model-agnostic and easily-constructed attack, called CommanderGabble, which uses fast speech to camouflage voice commands. Both humans and ASR systems often misinterpret fast speech, and such misinterpretation can be exploited to launch hidden voice command attacks. Specifically, by carefully manipulating the phonetic structure of a target voice command, ASRs can be caused to derive a hidden meaning from the manipulated, high-speed version. We implement the discovered attacks both over-the-wire and over-the-air, and conduct a suite of experiments to demonstrate their efficacy against 7 practical ASR systems. Our experimental results show that the over-the-wire attacks can disguise as many as 96 out of 100 tested voice commands into adversarial ones, and that the over-the-air attacks are consistently successful for all 18 chosen commands in multiple real-world scenarios. 

   more » « less
5. Noise schemas aid hearing in noise

   https://doi.org/10.1073/pnas.2408995121  

   Hicks, Jarrod_M; McDermott, Josh_H (November 2024, Proceedings of the National Academy of Sciences)

   Human hearing is robust to noise, but the basis of this robustness is poorly understood. Several lines of evidence are consistent with the idea that the auditory system adapts to sound components that are stable over time, potentially achieving noise robustness by suppressing noise-like signals. Yet background noise often provides behaviorally relevant information about the environment and thus seems unlikely to be completely discarded by the auditory system. Motivated by this observation, we explored whether noise robustness might instead be mediated by internal models of noise structure that could facilitate the separation of background noise from other sounds. We found that detection, recognition, and localization in real-world background noise were better for foreground sounds positioned later in a noise excerpt, with performance improving over the initial second of exposure to a noise. These results are consistent with both adaptation-based and model-based accounts (adaptation increases over time and online noise estimation should benefit from acquiring more samples). However, performance was also robust to interruptions in the background noise and was enhanced for intermittently recurring backgrounds, neither of which would be expected from known forms of adaptation. Additionally, the performance benefit observed for foreground sounds occurring later within a noise excerpt was reduced for recurring noises, suggesting that a noise representation is built up during exposure to a new background noise and then maintained in memory. These findings suggest that noise robustness is supported by internal models—“noise schemas”—that are rapidly estimated, stored over time, and used to estimate other concurrent sounds. 

   more » « less