-
Categories
-
Pharmaceutical Intermediates
-
Active Pharmaceutical Ingredients
-
Food Additives
- Industrial Coatings
- Agrochemicals
- Dyes and Pigments
- Surfactant
- Flavors and Fragrances
- Chemical Reagents
- Catalyst and Auxiliary
- Natural Products
- Inorganic Chemistry
-
Organic Chemistry
-
Biochemical Engineering
- Analytical Chemistry
- Cosmetic Ingredient
-
Pharmaceutical Intermediates
Promotion
ECHEMI Mall
Wholesale
Weekly Price
Exhibition
News
-
Trade Service
Source—Zhao Jizhong, editor—Wang Sizhen, Fang Yiyi, editor—Wang Sizhen
, more than 70 million people around the world are suffering from serious hearing problems, but the needs of the hearing-impaired community usually do not attract enough attention from the consumer electronics industry
。 With the vigorous development of artificial intelligence, in addition to the original interpersonal communication needs, the demand for human-computer interaction is also increasing[1], which puts forward higher requirements
for the auxiliary communication system suitable for hearing-impaired groups.
However, on the one hand, existing sign language interpretation systems based on image recognition [2] or sensor gloves [3] lack the conditions for practical application due to various technical limitations, and natural sign language itself is also due to unique grammar rules [4], It is not conducive to the use of sign language-based interpretation systems for human-computer interaction
by hearing-impaired groups.
On the other hand, unlike sign language, speech recognition by obtaining information from throat vibration is more direct and convenient, and does not require any professional training, and new progress has been made in the research of flexible wearable laryngeal vibration sensors [5].
Therefore, the development of speech recognition assisted communication systems without sign language interpretation may become the key to improving the
lives of hearing-impaired users and facilitating human-computer interaction.
January 28, 2022, Guo Wenxi/Wu Ronghui, School of Physical Science and Technology, Xiamen University The research group and collaborators published an online publication entitled " Self-powered speech recognition system for deaf users", which reports an interference-resistant speech recognition system
developed primarily for hearing-impaired users.
The system is powered by a self-powered triboelectric vibration sensor (STVS) that collects the signal with a softly woven nanofiber cellulose film (NFCF) as a vibration-sensitive layer to enable STVS High sensitivity
at a wide vibration frequency.
The context-based recognition model (CRM) can accurately identify a variety of common expressions and has the function
of voice recognition.
This speech recognition system can provide a convenient and efficient communication channel
for the hearing impaired, the hearing (non-hearing impaired group) and the Internet of Things.
The Guo Wenxi/Wu Ronghui research group has always been committed to the development of
flexible wearable sensors.
The study first started from the fact that many hearing-impaired people only have hearing loss and intact vocal ability [6], based on the assumption that "only a limited vocabulary can well cover the communication needs of a specific situation", and finally established a reliance on STVS and CRM of the speech recognition system (Figure 1).
STVS attaches to the surface of the skin near the vocal cords and uses the friction nanogenerator (TENG) principle [7] to achieve sono-electrical energy conversion
.
These electrical signals are then sent to the CRM for personalized training and recognition
.
Finally, the trained model is invoked to identify the voice signals of the hearing-impaired user in real time and convert them into voice or text commands for controlling the smart home
.
Figure 1 Conceptual diagram of a speech recognition system (Source: Zhao et al.
, CRPS, 2022).
NFCF is soft, comfortable and safe [8]； STVS is sensitive, accurate and efficient
.
As can be seen from Figure 2C, SVTS has a broadband response, showing resonance characteristics in the 228-291 Hz range, very close to the fundamental frequency
of the human voice.
At the single-frequency response at lower (227 Hz), medium (521 Hz), and higher (829 Hz) frequencies of vocals, only no more than 0.
5% drift was observed (Figure 2D).
。 This shows that STVS can accurately record vibration information with little distortion and can distinguish between multiple vibration components of different frequencies
.
In addition, STVS shows a high signal-to-noise ratio (Figure 2E) and durability over one million cycles (Figure 2G).
Therefore, it is reasonable to assume that STVS has excellent sensing performance in the main audible sound range.
Figure 2 Vibration acquisition and electrical signal output performance of STVS (Source: Zhaoet al.
, CRPS, 2022).
Hearing-impaired users need personalized training models
.
Most hearing-impaired users have difficulty distinguishing between noisy and quiet environments, so improve the speech recognition system anti-interference ability is necessary
.
As shown in Figure 3A, STVS is still able to record vocals
with excellent accuracy at over 90 dB of noise.
The researchers invited four hearing-impaired volunteers who suffered hearing loss due to a drug reaction to a speech recognition test
.
Volunteers have their own unique but repeatable way of pronouncing a word, so that their voice vibration and words can establish a one-to-one correspondence.
This is the basis for the establishment of the voice recognition system for hearing-impaired users, and the existing speech recognition system is difficult to meet the needs
of hearing-impaired users.
The researchers invited four hearing-impaired volunteers to build a dataset by repeating 17 words 80 times each, and then modeled it using a single hidden layer long short-term memory (LSTM) algorithm.
When people use their voice to control smart home systems, there is often a similar language sequence, such as "turn on the air conditioner in the bedroom"
.
Therefore, the above short sentences can be grouped according to the volunteers' language habits (Figure 3H).
The volunteers' recognition accuracy in a certain category of words increased by an average of 3.
0% to 92.
3%.
Figure 3 Speech recognition of hearing-impaired users by STVS (Source: Zhaoet al.
, CRPS, 2022)
The identification system works with security
.
In order to improve the security of the smart home system, the researchers captured the "voiceprint" from the voices of volunteers and set up an intelligent voice-controlled security system
accordingly.
For any user who accesses the security system, its voice spectrum will be carefully analyzed and compared with the registration password to determine whether it is an authorized user and then decide whether to unlock
it.
This process can effectively protect the smart home system from abuse
.
Article conclusion and discussion, inspiration and prospects
All authors agree that the key to helping hearing-impaired users ease communication difficulties is to enable them to communicate in the same way as hearing-impaired people, that is, by speaking
with their voices.
On the one hand, this can make communication between the hearing-impaired and non-hearing-impaired groups more convenient, and on the other hand, it will also make it easier for hearing-impaired users to interact
with the Internet of Things.
Therefore, the authors of this work introduce natural cotton wool cellulose in terms of materials, simple woven structures in terms of structure, and word order division in recognition models, so as to establish a speech recognition system
with good use effect.
Commands for hearing-impaired users can be converted from throat vibration to text or speech in real time for human-computer interaction
.
Along this path, there is still room
for improvement in increasing the upper limit of frequency response, using phonemes as the minimum identification unit, increasing recognition accuracy, and eliminating voice interference signals.
Original link: _mstmutation="1" _msthash="162189" _msttexthash="126475791">Zhao Jizhong of the School of Physical Science and Technology, Xiamen University is the first author of the work, and Professor Guo Wenxi is the final corresponding author Welcome to scan the code to join Logical Neuroscience Literature Study 2
.
Professor Guo Wenxi is mainly engaged in the research of soft matter and flexible electronic skin, and has been published in Adv.
Mater, JACS, Nano Lett and other journals published more than 80 SCI articles, H factor 40
.
First author: Zhao Jizhong (left); Corresponding author: Guo Wen (right) (photo courtesy of Guo Wenxi's research group).
for autism social disorder based on synaptic cell biological characteristics[3] Expert comments iScience—Li Yan's team revealed the molecular mechanism of familial epilepsy [4] Cell Death Discov—Kang Jiuhong's team found that NRG1 is expected to become a new target for the treatment of schizophrenia caused by intrauterine growth restriction [5] Nature—Zhang Shicheng et al.
analyzed the design principle of DREADD, a chemical genetic tool based on muscarinic acetylcholine receptors [6] eLife-Chen Shuyi's team first revealed the m6A epitranscriptional regulation mechanism of state transition between neural progenitor cells and glial cells [7] Nature-Shi Songhai's research group revealed a new mechanism regulating the spatial fine structure arrangement and loop assembly of neurons in the neocortex of the brain [8] Mol Psychiatry—Zhang Jie's group revealed the association between morphological differentiation of the cortex and subcortical regions and children's cognitive function and psychiatric diseases [9] NeuroImage—Yan Chaogan's team developed the Think-Aloud fMRI research paradigm and portrayed the brain representation model of resting spontaneous thinking [10] Mol Psychiatry—Chen Yu et al.
studied changes in immune-related genes in the brains of patients with mental disorders and neurodegenerative diseases across diseases, NeuroAI Reading Club [1] NeuroAI Reading Club Launched—Exploring the Frontier Intersection
of Neuroscience and Artificial Intelligence Recommended for High-quality Scientific Research Training Courses [1] Symposium on Patch-Clamp and Optogenetics and Calcium Imaging Technology (January 7-8, 2023 Tencent Meeting) [2].
The 10th NIR Training Camp (Online: 2022.
11.
30~12.
20) [3] The 9th EEG Data Analysis Flight (Training Camp: 2022.
11.
23-12.
24) Welcome to "Logical Neuroscience" [1] " Logical Neuroscience "Recruitment for Editorial/Operation Positions (Online Office)[2]" Logical Neuroscience "Candidate for Associate Editor/Editor/Operation Position (Online Office)[3] Recruitment—" Logical Neuroscience "Job Interpretation/Writing Position ( Online Part-time, Online Office) References (Swipe Up and Down to Read)
[1] Guo, H.
, Pu, X.
, Chen, J.
, Meng, Y.
, Yeh, M.
H.
, Liu, G.
, Tang, Q.
, Chen, B.
, Liu, D.
, Qi, S.
, et al.
(2018).
A highly sensitive, self-powered triboelectric auditory sensor for social robotics and hearing aids.
Sci Robot3, eaat2516.
10.
1126/scirobotics.
aat2516.
[2] Rajam, P.
S.
, and Balakrishnan, G.
(2012).
Recognition of Tamil Sign Language Alphabet using Image Processing to aid Deaf-Dumb People.
Procedia Engineering30, 861-868.
10.
1016/j.
proeng.
2012.
01.
938.
[3] Zhou, Z.
, Chen, K.
, Li, X.
, Zhang, S.
, Wu, Y.
, Zhou, Y.
, Meng, K.
, Sun, C.
, He, Q.
, Fan, W.
, et al.
(2020).
Sign-to-speech translation using machine-learning-assisted stretchable sensor arrays.
Nature Electronics3, 571-578.
10.
1038/s41928-020-0428-6.
[4] Hofer, T.
(2017).
Is Lhasa Tibetan Sign Language emerging, endangered, or both? Int J Soc Lang2017, 113-145.
10.
1515/ijsl-2017-0005.
[5] Dinh Le, T.
S.
, An, J.
, Huang, Y.
, Vo, Q.
, Boonruangkan, J.
, Tran, T.
, Kim, S.
W.
, Sun, G.
, and Kim, Y.
J.
(2019).
Ultrasensitive Anti-Interference Voice Recognition by Bio-Inspired Skin-Attachable Self-Cleaning Acoustic Sensors.
ACS Nano13, 13293-13303.
10.
1021/acsnano.
9b06354.
[6] Fu, S.
, Chen, G.
, Dong, J.
, and Zhang, L.
(2010).
Prevalence and etiology of hearing loss in primary and middle school students in the Hubei Province of China.
Audiol Neurootol15, 394-398.
10.
1159/000307346.
[7] Fan, F.
-R.
, Tian, Z.
-Q.
, and Lin Wang, Z.
(2012).
Flexible triboelectric generator.
Nano Energy1, 328-334.
10.
1016/j.
nanoen.
2012.
01.
004.
[8] Lin, C.
, Wang, Q.
, Deng, Q.
, Huang, H.
, Huang, F.
, Huang, L.
, Ni, Y.
, Chen, L.
, Cao, S.
, and Ma, X.
(2019).
Preparation of highly hazy transparent cellulose film from dissolving pulp.
Cellulose26, 4061-4069.
10.
1007/s10570-019-02367-3.
End of article
