Unlock instant, AI-driven research and patent intelligence for your innovation.

A sign language recognition method based on commercial Wi-Fi

A recognition method, sign language technology, applied in the field of sign language recognition, achieves the effect of low price, reduced difficulty and cost saving

Active Publication Date: 2019-06-18
TIANJIN UNIV
View PDF7 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] The present invention provides a sign language recognition method based on commercial Wi-Fi. The present invention is used to recognize continuous sign language sentences. The problem of automatically extracting sign language word features avoids inappropriate features manually selected empirically, and integrates a series of noise reduction and partial multipath methods to reduce the impact of noise. See the description below for details:

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A sign language recognition method based on commercial Wi-Fi
  • A sign language recognition method based on commercial Wi-Fi
  • A sign language recognition method based on commercial Wi-Fi

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0050] A kind of sign language recognition method based on commercial Wi-Fi, see figure 1 , the method includes the following steps:

[0051] 101: In an environment where a pair of wireless transceivers are deployed, the subject performs sign language actions, and the original CSI sequence is collected by the receiver;

[0052] 102: Perform a series of denoising processes on the original trajectory of the CSI sequence, including: partial multipath denoising, outlier filtering, bandpass filtering and locally weighted linear regression;

[0053] 103: After removing the noise, use the sign language sentence segmentation component to segment the filtered CSI sequence, and separate the sign language words;

[0054] That is, a segmentation method based on PCA (Principal Component Analysis) is used to find the beginning and end of sign language words, and then the segmented sign language words are forwarded to the sign language recognition component, and step 104 is executed. Among...

Embodiment 2

[0060] The scheme in embodiment 1 is further introduced below in conjunction with specific examples and calculation formulas, see the following description for details:

[0061] 1. Using commercial Wi-Fi to collect CSI signals

[0062] The present invention has used two notebook computers that have Intel 5300 network card and Ubuntu system. One of the laptop's antennas acts as a transmitter, operating at 5.825GHz in IEEE 802.11n monitor mode, and the other laptop's antenna acts as a receiver. During the measurement, the transmitter sends about 1000 data packets per second to the receiver via the WiFi router. The present invention helps realize higher CSI sampling rate by setting a higher sending packet frequency, ensures the time resolution of the CSI value, captures subtle changes in the CSI stream, and maximizes the details of different sign language movements. Transceivers are placed at a height of 1.5m and the distance between them is 80cm. Using one antenna as the tran...

Embodiment 3

[0122] Attached below Figure 6 The functions and effects in the above-mentioned embodiments 1 and 2 are demonstrated.

[0123] This example uses CSI data processing as an example to give a specific implementation method, and the specific steps are as follows:

[0124] Use a laptop as the Wi-Fi access point, that is, the sender, and another laptop as the receiver. Both laptops are installed with Intel 5300 NIC and Ubuntu 14.04 LTS desktop system. The transmitter has 1 antenna. Each There are 3 antennas at the receiving end, the distance between the 3 antennas at each end is one wavelength (5.79cm), and they are located on a straight line. The transmitting end and the receiving end are placed on two small table boards, 1.5m above the ground, and the distance between the transmitting end and the receiving end is 80cm. During the measurement, the transmitter sends about 1000 data packets per second to the receiver via the WiFi router. Setting a higher sending packet frequency ...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a sign language recognition method based on commercial Wi-Fi, which comprises the following steps: sequentially carrying out partial multi-path denoising, outlier filtering, band-pass filtering and local weighted linear regression processing on an original track of a CSI sequence; Segmenting the filtered CSI sequence by using a sign language sentence segmentation component,and separating sign language words; Constructing a deep belief network model for extracting fine-grained characteristics of the sign language words, processing the fine-grained characteristics by utilizing the improved hidden Markov model, and identifying the sign language words; screening and correcting the recognized correct sign language words and similar sign language words based on an N-grammodel, so that the recognition precision is improved; And combining all the corrected sign language words together to reconstruct sign language sentences, and outputting the sign language sentences to a smart phone or other interactive equipment through a voice assistant. According to the method, inappropriate features which are manually selected empirically are avoided, and a series of noise reduction and partial multi-path methods are integrated, so that the influence of noise is reduced.

Description

technical field [0001] The invention relates to the field of sign language recognition, in particular to a method for recognizing sign language based on commercial Wi-Fi. Background technique [0002] As the common language of the deaf community, sign language is also an important bridge for communication between the deaf and normal people. However, normal people without special training cannot understand sign language. Communication barriers between the two groups remain. As an important group in every country, deaf people should have a more convenient way to communicate with normal people. If sign language can be converted into speech through recognition technology, it will greatly promote the communication between deaf-mute and normal people. [0003] Existing sign language recognition methods can be divided into two categories: device-based recognition and device-free recognition. Device-based sign language recognition methods for signals include vision-based and sen...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
IPC IPC(8): G06K9/00H04W4/30
CPCY02D30/70
Inventor 张翼翔张蕾阮新
Owner TIANJIN UNIV