Lead AiSpeak Portal productProduct Name: AiSpeak Portal Description: AiSpeak Portal is a voice-controlled internet service that provides news, broadcasting, video, intelligent control and social networking. Individuals can use simple voice commands to regain their activities on the internet. In addition, the AiSpeak Portal enables patients to easily communicate and express their needs to others, by sending messages to designated mobile devices when their family members or health caregivers are out of reach. Product release:The download link of the Android tablet is as follows:https://play.google.com/store/apps/details?id=com.aprevent.aispeakProvide the main account user to useThe Android mobile download link is as follows:https://play.google.com/store/apps/details?id=com.aprevent.aispeakOnly available for chat and messaging with the main account user - Patent publication.Patent pending, Method of voice input operation.

- Zenbo Robot Speaker Identification System Design, Da Vinci Innovation Lab1. Design and implementation of initial text-dependent speaker enrollment system: Prompt the user to repeat a given trigger word five times and then use these samples to enroll the user’s voice.2. Design and implementation of text-dependent speaker identification testing and speaker incremental enrollment system: The user can begin to say trigger word to wake the robot up. This trigger word sample will be used to enroll the user’s voice again. Next time the user speaks the trigger word, the text-dependent speaker identification recognition rate can be improved.3. Design and implementation of text-independent speaker identification testing and speaker incremental enrollment system: After the robot is woken up, the user can begin to say command word to command robot. This command word sample will be used to enroll the user’s voice again. Next time the user speaks the command word, the text-independent speaker identification recognition rate can be improved.4. Write the test program for QTC to test and produce Speaker & Impostor recognition rate.5. Integrate all the code associated with the Speaker ID function, such as integrated Voice Activity Detection, Speech Recognition, Background Noise Detection, Utterance Verification, etc. 6. Debug Robot with Speaker ID related issues.- Patent publication.1. U.S. Patent 20190096409, Electronic apparatus having incremental enrollment unit and method thereof.

- Foxconn sent me to the Massachusetts Institute of Technology (MIT) Computer Science and Artificial Intelligence Laboratory Spoken Language Systems Group as Visiting Scientist four years (03/2010-12/2013). Focus: speaker recognition, language recognition and emotion recognition.- Designed a speech/speaker recognition system capable of recognizing spoken commands, identifying the user via a cloud-based speaker recognition system, and then using these commands to control the movements of a robot.- Designed a speaker diarization system capable of searching for a desired section from a number of stored video/audio files. Then the user could know who spoken when.- Patent publication.1. U.S. Patent 20160351185, Voice recognition device and method.2. U.S. Patent 20150088513, Sound processing system and related method.

- Developed a speaker diarization system that could process streaming audio from multiple speakers.1. This system was able to take an audio stream and automatically identified segments of speech from the same speaker. 2. Incorporated speaker identification into this system to identify the talkers. 3. Identified both known and unknown speakers.- Developed and incorporated a language identification capability into the streaming system, and developed a multi-language speaker identification capability.1. Detected whether multiple people were speaking simultaneously 2. If so, identified the simultaneous speakers.3. The challenges of this task were significant, as segments containing simultaneous speech were often very short; so short, in fact, that it would be rather difficult to characterize even just one speaker, much less two or more.- Developed an ability to identify emotions and incorporated this capability into the streaming classification system.

- Designed Print Server, Router and VOIP hardware circuit board for OEM and ODM customers.1. According to the needs of the product, we used OrCAD tool to design with the required circuit functions. Looking for components in addition to the main IC, the design of other passive components must be considered together with the second source into the circuit diagram. Cost down was the most basic work of hardware engineers.2. Verified that the component had the correct position and height on the circuit board to prevent the component from blocking the housing. Power, clock and ground must be properly routed in the layout so that EMC could easily pass certification. Layout was confirmed carefully before they produced a Gerber file.3. During the PCB production from Gerber, we should be ready in advance of the required components.4. After the PCB production was complete, we had to solder manually two samples by ourselves. After the soldering was complete, we verified that the sample was working properly.5. We could start the sample run processing if two samples of hand soldering SMT work.6. We verified that each piece sample was working properly after sample run processing was complete.7. Samples could be transferred to the software engineer after samples were confirmed to be error-free.8. We could go to do EMC certification after software engineers designed software that could do EMC certification.9. According to the product category, we helped our customers to get the certification they needed.10. We did pilot run to confirm that the product had high yield rate before the product massed production.- Products passed EMC, Safety and Telecom Certification.