EERA-ASR: An Energy-Efficient Reconfigurable Architecture for Automatic Speech Recognition With Hybrid DNN and Approximate Computing

Liu, Bo; Hai, Qin; Gong, Yu; Ge, Wei; Xia, Mengwen; Shi, Longxing

doi:10.1109/access.2018.2870273

Cited by 40 publications

(25 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…On the one hand, in DNNs, the numbers of multiplication and addition operations required for the neural network layers such as convolution layers and fully connected layers are almost the same, but the power consumption of multiplication operation accounts for up to 96% of the total power consumption [9]. In our previous works [10], we have proposed a DNN network for speech recognition and a DNN accelerator architecture with approximate multiplication units to process different layers of the DNN. Implemented under 28nm CMOS technology, the power consumption of this work is 53.7 mW and the energy efficiency is 3.3 TOPS/W.…”

Section: Preliminaries a Energy Efficient Kws System Based On Bwnmentioning

confidence: 99%

EERA-KWS: A 163 TOPS/W Always-on Keyword Spotting Accelerator in 28nm CMOS Using Binary Weight Network and Precision Self-Adaptive Approximate Computing

Liu

Wang²,

et al. 2019

IEEE Access

Self Cite

View full text Add to dashboard Cite

This paper proposed an energy-efficient reconfigurable accelerator for keyword spotting (EERA-KWS) based on binary weight network (BWN) and fabricated in 28-nm CMOS technology. This keyword spotting system consists of two parts: the feature extraction based on melscale frequency cepstral coefficients (MFCC) and the keywords classification based on a BWN model, which is trained through the Google's Speech Commands database and deployed on our custom. To reduce the power consumption while maintaining the system recognition accuracy, we first optimize the MFCC implementation with approximate computing techniques, including Pre-emphasis coefficient transformation, rectangular Mel filtering, Framing and FFT optimization. Then, we propose a precision self-adaptive reconfigurable accelerator with digital-analog mixed approximate computing units to process the BWN efficiently. Based on the SNR prediction of background noise and post-detection of network output confidence, the BWN accelerator data path can be dynamically and adaptively reconfigured as 4, 8, or 16 bits. For the BWN accelerator, we proposed a time-delay based addition unit to process bit-wise approximate computing for the convolution layers and fully connected layers, and a LUT based unit for the activation layers. Implemented under TSMC 28 nm HPC+ process technology, the estimated power is 77.8 µW ∼ 115.9µW, the energy efficiency can achieve 163 TOPS/W, which is over 1.8× better than the state-of-the-art architecture.

show abstract

Section: Preliminaries a Energy Efficient Kws System Based On Bwnmentioning

confidence: 99%

EERA-KWS: A 163 TOPS/W Always-on Keyword Spotting Accelerator in 28nm CMOS Using Binary Weight Network and Precision Self-Adaptive Approximate Computing

Liu

Wang²,

et al. 2019

IEEE Access

Self Cite

View full text Add to dashboard Cite

show abstract

“…Automatic recognition and identification of emotions from speech signals in speech emotion recognition (SER) using machine learning is a challenging task [1]. SER is a quick and usual method of communication and exchanging information among humans and computers and has many real world applications in the domain of Human-computer interaction (HCI).…”

Section: Introduction Of Sermentioning

confidence: 99%

Clustering-Based Speech Emotion Recognition by Incorporating Learned Features and Deep BiLSTM

2020

View full text Add to dashboard Cite

Emotional state recognition of a speaker is a difficult task for machine learning algorithms which plays an important role in the field of speech emotion recognition (SER). SER plays a significant role in many real-time applications such as human behavior assessment, human-robot interaction, virtual reality, and emergency centers to analyze the emotional state of speakers. Previous research in this field is mostly focused on handcrafted features and traditional convolutional neural network (CNN) models used to extract high-level features from speech spectrograms to increase the recognition accuracy and overall model cost complexity. In contrast, we introduce a novel framework for SER using a key sequence segment selection based on redial based function network (RBFN) similarity measurement in clusters. The selected sequence is converted into a spectrogram by applying the STFT algorithm and passed into the CNN model to extract the discriminative and salient features from the speech spectrogram. Furthermore, we normalize the CNN features to ensure precise recognition performance and feed them to the deep bi-directional long short-term memory (BiLSTM) to learn the temporal information for recognizing the final state of emotion. In the proposed technique, we process the key segments instead of the whole utterance to reduce the computational complexity of the overall model and normalize the CNN features before their actual processing, so that it can easily recognize the Spatio-temporal information. The proposed system is evaluated over different standard dataset including IEMOCAP, EMO-DB, and RAVDESS to improve the recognition accuracy and reduce the processing time of the model, respectively. The robustness and effectiveness of the suggested SER model is proved from the experimentations when compared to state-of-the-art SER methods with an achieve up to 72.25%, 85.57%, and 77.02% accuracy over IEMOCAP, EMO-DB, and RAVDESS dataset, respectively. INDEX TERMS Speech emotion recognition, deep bidirectional long shot term memory, key segment sequence selection, normalization of CNN features, radial-based function network (RBFN).

show abstract

“…Speech interaction has become an essential way of humanmachine interaction [1], [2], in which, automatic speech recognition (ASR) plays a vital role in perceiving speech signals. In scenarios such as energy-constrained, network restricted wearable devices, energy-efficient speech recognition is important for the working and standby time of the devices.…”

Section: Introductionmentioning

confidence: 99%

MSP-MFCC: Energy-Efficient MFCC Feature Extraction Method With Mixed-Signal Processing Architecture for Wearable Speech Recognition Applications

Yang

Lan

et al. 2020

IEEE Access

View full text Add to dashboard Cite

Feature extraction is an essential part of automatic speech recognition (ASR) to compress raw speech data and enhance features, where conventional implementation methods based on the digital domain have encountered energy consumption and processing speed bottlenecks. Thus, we propose a Mixed-Signal Processing (MSP) architecture to efficiently extract Mel-Frequency Cepstrum Coefficients (MFCC) features. We design MSP-MFCC to pre-process speech signals in the analog domain, which significantly reduces the cost of the analog-to-digital converter (ADC), as well as the computational complexity of the digital backend. Moreover, MSP-MFCC eliminates the time-consuming Fourier transform in the conventional digital realization by improving processing flow. We fabricated the analog part based on 180nm CMOS mixedsignal technology, then measured the chip. The measured results show the energy consumption of MSP-MFCC is 0.72 µJ/frame, and the processing speed is up to 45.79 µs/frame. MSP-MFCC achieves 95% energy saving and about 6.4× speedup than state of the art. Further, by using the features extracted by MSP-MFCC, speech recognition simulation reaches the accuracy of 98.2%, which also keeps the leading performance to its current counterparts. The proposed MFCC extractor is competitive for integration in the ultra-low-power always-on wearable speech recognition applications. INDEX TERMS Mixed signal processing architecture, energy-efficient feature extraction, mel-frequency cepstrum coefficients (MFCC), wearable speech recognition application.

show abstract

EERA-ASR: An Energy-Efficient Reconfigurable Architecture for Automatic Speech Recognition With Hybrid DNN and Approximate Computing

Cited by 40 publications

References 16 publications

EERA-KWS: A 163 TOPS/W Always-on Keyword Spotting Accelerator in 28nm CMOS Using Binary Weight Network and Precision Self-Adaptive Approximate Computing

EERA-KWS: A 163 TOPS/W Always-on Keyword Spotting Accelerator in 28nm CMOS Using Binary Weight Network and Precision Self-Adaptive Approximate Computing

Clustering-Based Speech Emotion Recognition by Incorporating Learned Features and Deep BiLSTM

MSP-MFCC: Energy-Efficient MFCC Feature Extraction Method With Mixed-Signal Processing Architecture for Wearable Speech Recognition Applications

Contact Info

Product

Resources

About