Authentication based on correlation of multiple pulse signals
Techniques are described for user authentication based on correlation of multiple pulse signals that each describe an independently determined pulse pattern for an individual. Different measurements of an individual's pulse can be generated in response to a request from the individual to access a computing system, such as a kiosk. The different measurements can be generated independently of one another using different types of sensor devices. For example, a measurement can be made using a radar sensor to detect movements caused by the user's pulse, and another measurement can be made through analysis of video data. If the independently generated pulse patterns are sufficiently similar to each other (e.g., determined using machine learning or other techniques), a determination may be made that the individual is not attempting to spoof another individual, and access to the computing system can be granted based at least partly on the determination.
1. A computer-implemented method performed by at least one processor, the method comprising:
accessing, by the at least one processor, first sensor data generated by a first sensor device that is included in a kiosk, the first sensor data describing at least one first physiological characteristic of an individual;
accessing, by the at least one processor, second sensor data generated by a second sensor device in proximity to the kiosk, the second sensor device included in a wearable device worn or carried by the individual, the second sensor data generated independently of the first sensor data, the second sensor data describing at least one second physiological characteristic of the individual, wherein the first sensor data and the second sensor data are accessed in response to a request of the individual to access information through the kiosk;
providing, by the at least one processor, the first sensor data as input to a first classifier that is trained to predict pulse patterns based on a type of the first sensor device, and obtaining a first pulse pattern as output of the first classifier;
providing, by the at least one processor, the second sensor data as input to a second classifier, different from the first classifier, that is trained to predict pulse patterns based on a type of the second sensor device, and obtaining a second pulse pattern as output of the second classifier; and
determining, by the at least one processor, a degree of correspondence between the first pulse pattern and the second pulse pattern and, based at least partly on the degree of correspondence, determining a result of the request of the individual to access information through the kiosk.
2. The method of claim 1 , wherein:
the first sensor device is at least one of a radar sensor device, a laser Doppler measurement device, and a laser vibrometer;
the first sensor data indicates movement of at least one body part of the individual, the movement associated with a pulse of the individual; and
analyzing the first sensor data includes extracting the first pulse pattern based on the movement indicated in the first sensor data.
3. The method of claim 1 , wherein:
the first sensor device is a camera;
the first sensor data includes video of at least one body part of the individual; and
analyzing the first sensor data includes analyzing the video to determine the first pulse pattern based on one or more of a movement and a skin color fluctuation shown in the video.
4. The method of claim 1 , wherein the determined result is a denial of access to the kiosk, based on an absence of a pulse in at least one of the first pulse pattern and the second pulse pattern.
5. The method of claim 1 , wherein the determined result is a denial of access to the kiosk based on a difference, between the first pulse pattern and the second pulse pattern, in one or more of a frequency and a time offset of a pulse.
6. The method of claim 1 , wherein the determined result is a grant of access to the kiosk based on the first pulse pattern being within a threshold degree of similarity to the second pulse pattern in both a frequency and a time offset of a pulse.
7. The method of claim 1 , wherein the first classifier and the second classifier include one or more of a neural network, a recurrent neural network (RNN), a long short-term memory (LSTM) network, or a support vector machine (SVM).
8. A system comprising:
at least one processor; and
a non-transitory memory communicatively coupled to the at least one processor, the non-transitory memory storing instructions which, when executed, cause the at least one processor to perform operations comprising:
accessing first sensor data generated by a first sensor device that is included in a kiosk, the first sensor data describing at least one first physiological characteristic of an individual;
accessing second sensor data generated by a second sensor device in proximity to the kiosk, the second sensor device included in a wearable device worn or carried by the individual, the second sensor data generated independently of the first sensor data, the second sensor data describing at least one second physiological characteristic of the individual, wherein the first sensor data and the second sensor data are accessed in response to a request of the individual to access information through the kiosk;
providing the first sensor data as input to a first classifier that is trained to predict pulse patterns based on a type of the first sensor device, and obtaining a first pulse pattern as output of the first classifier;
providing the second sensor data as input to a second classifier, different from the first classifier, that is trained to predict pulse patterns based on a type of the second sensor device, and obtaining a second pulse pattern as output of the second classifier; and
determining a degree of correspondence between the first pulse pattern and the second pulse pattern and, based at least partly on the degree of correspondence, determining a result of the request of the individual to access information through the kiosk.
9. The system of claim 8 , wherein:
the first sensor device is at least one of a radar sensor device, a laser Doppler measurement device, and a laser vibrometer;
the first sensor data indicates movement of at least one body part of the individual, the movement associated with a pulse of the individual; and
analyzing the first sensor data includes extracting the first pulse pattern based on the movement indicated in the first sensor data.
10. The system of claim 8 , wherein:
the first sensor device is a camera;
the first sensor data includes video of at least one body part of the individual; and
analyzing the first sensor data includes analyzing the video to determine the first pulse pattern based on one or more of a movement and a skin color fluctuation shown in the video.
11. The system of claim 8 , wherein the determined result is a denial of access to the kiosk, based on an absence of a pulse in at least one of the first pulse pattern and the second pulse pattern.
12. The system of claim 8 , wherein the determined result is a denial of access to the kiosk based on a difference, between the first pulse pattern and the second pulse pattern, in one or more of a frequency and a time offset of a pulse.
13. The system of claim 8 , wherein the determined result is a grant of access to the kiosk based on the first pulse pattern being within a threshold degree of similarity to the second pulse pattern in both a frequency and a time offset of a pulse.
14. One or more non-transitory, computer-readable storage media storing instructions which, when executed, cause at least one processor to perform operations comprising:
accessing first sensor data generated by a first sensor device that is included in a kiosk, the first sensor data describing at least one first physiological characteristic of an individual;
accessing second sensor data generated by a second sensor device in proximity to the kiosk, the second sensor device included in a wearable device worn or carried by the individual, the second sensor data generated independently of the first sensor data, the second sensor data describing at least one second physiological characteristic of the individual, wherein the first sensor data and the second sensor data are accessed in response to a request of the individual to access information through the kiosk;
providing the first sensor data as input to a first classifier that is trained to predict pulse patterns based on a type of the first sensor device, and obtaining a first pulse pattern as output of the first classifier;
providing the second sensor data as input to a second classifier, different from the first classifier, that is trained to predict pulse patterns based on a type of the second sensor device, and obtaining a second pulse pattern as output of the second classifier; and
determining a degree of correspondence between the first pulse pattern and the second pulse pattern and, based at least partly on the degree of correspondence, determining a result of the request of the individual to access information through the kiosk.
15. The computer-implemented method of claim 1 , wherein the first sensor device employs a first sensing method, and wherein the second sensor device employs a second, different sensing method.