DeepEar: robust smartphone audio sensing in unconstrained acoustic environments using deep learning

By Nicholas Donald Lane, Petko Georgiev, Lorena Qendro. Ubiquitous Computing, 2015

335 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Signal Processing

View Paper: https://www.semanticscholar.org/paper/9cdd742261c9f07887354622f79ade8226d625c7

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Signal Processing

Citations in the corpus

Learning IoT in Edge: Deep Learning for the Internet of Things with Edge Computing (2018, methodology): “[10] proposed new acceleration engines, such as DeepEar and DeepX, to support different deep learning applications in latest mobile system-on-chips (SoC).”
DeepX: A Software Accelerator for Low-Power Deep Learning Inference on Mobile Devices (2016, background): “mizations such as [33] that scaled down DNNs to run directly on a DSP only, offering energy efficiency.”
Deep Learning for Sensor-based Activity Recognition: A Survey (2017, methodology): “Although some existing work adopted deep HAR on smartphone (Lane et al., 2015) and watch (Bhattacharya and Lane, 2016), they are still far from online and mobile deployment.”
Efficient Acceleration of Deep Learning Inference on Resource-Constrained Edge Devices: A Review (2023, methodology): “DeepEar [397] is DL-based audio sensing using a smartphone that used stacked RBM.”
DeepMon: Mobile GPU-based Deep Learning Framework for Continuous Vision Applications (2017, methodology): “took important first steps towards the real-time execution of DNN and CNN on mobile devices [35, 36].”
An Early Resource Characterization of Deep Learning on Wearables, Smartphones and Internet-of-Things Devices (2015, neutral): “Although preliminary in nature, the investigation reported here provides insights into a number of key areas.”
Deep Learning for RFID-Based Activity Recognition (2016, methodology): “Examples include the Convolutional Neural Network (CNN), widely used for image classification [13] and recently for speech recognition [24], the Deep Neural Network (DNN), used for speech recognition and audio sensing [25], and a multimodal structure used for audio-visual speech recognition [26].”
Accelerating Mobile Audio Sensing Algorithms through On-Chip GPU Offloading (2017, background): “A large body of research has been devoted to the use of heterogeneous computation via low-power co-processors [51, 45, 27, 40, 42, 54] and custombuilt peripheral devices [56] to accelerate or sustain power efficient processing for extended periods of time.”
DeepCham: Collaborative Edge-Mediated Adaptive Deep Learning for Mobile Object Recognition (2016, background): “Thus an edge-based solution is required which ideally would perform adaptation training on the mobile phones.”
Deep Learning in Human Activity Recognition with Wearable Sensors: A Review on Advances (2021, methodology): “[159] utilizes the smartphone’s digital signal processor (DSP) to reduce power consumption and improve inference time on smartphones.”
DeepSense: A Unified Deep Learning Framework for Time-Series Mobile Sensing Data Processing (2016, methodology): “DeepEar [22] uses Deep Boltzmann Machines to improve the performance of audio sensing tasks in an environment with background noise.”
Deep, Convolutional, and Recurrent Models for Human Activity Recognition Using Wearables (2016, positive): “…the use of pre-trained, fully-connected networks for automatic assessment in Parkinson’s Disease [ Hammerla et al. , 2015 ] , as emission model in Hidden Markov Models (HMMs) [ Zhang et al. , 2015; Alsheikh et al. , 2015 ] , and to model audio data for ubicomp applications [ Lane et al. , 2015 ] .”
MCDNN: An Approximation-Based Execution Framework for Deep Stream Processing Under Resource Constraints (2016, background): “For high-value scenarios such as speech and object recognition, researchers have crafted efficient DNNs by hand [32, 42, 45] or structured them for execution on efficient co-processors such as DSPs (for smaller DNNs) [31].”