WO2004042643A1

WO2004042643A1 - Fingerprint sensor and method for operating a fingerprint sensor

Info

Publication number: WO2004042643A1
Application number: PCT/DE2003/003541
Authority: WO
Inventors: Josef Riegebauer; Klaus Strohmayer
Original assignee: Infineon Technologies Ag
Priority date: 2002-11-05
Filing date: 2003-10-24
Publication date: 2004-05-21
Also published as: TW200407793A; DE10251460A1

Abstract

A fingerprint sensor and a circuit (5) for data compression are integrated on a semiconductor chip (1). According to the invention, the data values in a predetermined sequence, which are recorded and digitized by the sensor elements (2), are replaced section-by-section, e.g. line-by-line, with the exception of the data value of the first sensor element of a respective section, by the difference between the data value of the respective sensor element and by the sum of the data value of the sensor element, which is first in the relevant section, as well as by the differences by which the data values of the sensor element that follows the first sensor element were replaced.

Description

description

Fingerprint sensor and method for operating a fingerprint sensor

The present invention relates to a fingerprint sensor with data compression and a corresponding method for operating a fingerprint sensor.

An electronic fingerprint sensor captures a rasterized image of a fingerprint, i. H. the position of the furrows and ridges of the skin surface of the fingertip. The individual pixels are usually present in a rectangular grid and are detected by sensor elements arranged in rows and columns. In the case of an embodiment which can be produced inexpensively and with little wear, the fingerprint sensor has no moving parts; the sensor elements are essentially formed by conductor areas on the upper side of a semiconductor chip. These conductor surfaces are arranged in or just below an insulating contact surface for a finger. The electrical capacitance of a respective conductor surface with respect to the environment is measured in each case; these capacities change when a finger is placed according to the furrows and ridges of the skin surface (see M. Tartagni and R. Guerrieri, "A 390dpi Live

Fingerprint Imager Based on Feedback Capacitive Sensing Scheme "in 1997 IEEE International Solid-State Circuits Conference / Session 12 / Sensors / Paper FP 12.3, pages 200 and 201). A surface sensor records a two-dimensional image in contrast to a strip sensor, in which the finger has to be passed over a strip-shaped arrangement of sensor elements, each of which only detects one strip of the image. The sum of the strip images is usually fed to a surface image reconstruction.

The data recorded by the sensor elements are recorded and digitized using a connected circuit can then be further processed in a corresponding evaluation unit, e.g. B. to identify a person or to verify an access authorization. The evaluation circuit can be accommodated in a so-called host system (usually micro-controller, computer and the like). An integration of the evaluation circuit in a fingerprint sensor designed as a semiconductor chip would result in a very complex structure of the semiconductor chip.

To achieve the resolution required for personal identification, data formats of 8 bits per pixel are common. The amount of data per image of a two-dimensional arrangement of the sensor elements with 256 x 256 pixels is therefore approximately 64 kbytes. When capturing an image by means of a strip-shaped arrangement of sensor elements over which the finger is swiped, considerably higher data rates have to be managed, since the image has to be composed of a large number of individual image strips and these image strips must therefore partially overlap. Acceptable data transfer times therefore require parallel interfaces at the output of the fingerprint sensor. Using a serial interface significantly extends data acquisition and transmission.

WO 01/01327 AI describes an arrangement and a method for recording fingerprint data, in which the digitized data is compressed by a run length coding. For this purpose, the fingerprint sensor is combined with a data compression circuit to form a module which is connected to a data processing unit via a serial interface. The data processing unit decompresses the compressed fingerprint data received via the serial interface in a program-controlled manner. The run length coding, with which it is specified how many pixels have the same degree of brightness, is, according to this document, in the case of fingerprint data They are considered to be particularly efficient because there is an accumulation of equivalent data values in a fingerprint image.

The object of the present invention is to provide a practicable way of achieving a sufficient data transmission rate in the case of fingerprint sensors of high image resolution.

This object is achieved with the fingerprint sensor with the features of claim 1 and with the method for operating a fingerprint sensor with the features of claim 8. Refinements result from the dependent claims.

The fingerprint sensor is integrated in a semiconductor chip together with an electronic circuit, which is provided for a reduction or compression of the data, and with electrical connections, at which the data are output in the reduced or compressed form. The data compression is therefore already carried out in the semiconductor chip of the actual fingerprint sensor. The semiconductor chip has an arrangement of sensor elements, for. B. from conductor surfaces, for recording pixels on an upper side, which is provided as a contact surface for a finger. The arrangement of the sensor elements is not fixed and can be a matrix, a strip of a few, at least two, lines or a single line. Furthermore, a control circuit for recording and digitizing the data recorded by the sensor elements is integrated in the chip. In addition, there is the circuit which compresses the digitized data, which are referred to below as data values. Due to the complete integration of these circuits in the semiconductor chip of the fingerprint sensor, the data values can be quickly and efficiently brought into a compressed form and thus output. A preferred embodiment provides that the data values recorded and digitized by the sensor elements in sections in a predetermined sequence, e.g. B. line by line, with the exception of the data value of the first sensor element of a respective section

Difference between the data value of the respective sensor element and the sum of the data value of the first sensor element in the relevant section and the differences by which the data values of the sensor elements following the first sensor element were replaced. In principle, the difference between the data values of two successive pixels is always transmitted in this way, but it is achieved that at least almost all of the data values provide the original digitized data values of the relevant sensor elements until the end of such a section. So there is no cumulative error within such a section. Such a deterministic difference coding is particularly well adapted to the image structure of a typical fingerprint, since the typical wave pattern of a fingerprint caused by the furrows and ridges can be compressed with a deterministic difference coding without significant data loss.

The differences between the data values can be coded with a smaller number of bits than the data values themselves. B. coded with 8 bits per pixel, a respective difference z. B. can be encoded with 5 bits per pixel with sufficient accuracy. If the number of data values in a section is designated n and p and d are the number of bits per data value or per difference, (pd) (nl) / (pn) x 100% is the compression rate of the data values, ie that in percent specified saved portion of data bits per number of output data bits. In the given example with d = 5, p = 8 and n = 256, this compression rate is 37.35%. Low-pass filtering of the already digitized data values is preferably carried out before the data compression. A simple low-pass filter, e.g. B. a one-dimensional second-order low-pass filter. A second-order filter uses a linear combination of the current n-th data value x _n and the two previous data values x _n -ι and x _n - ₂ with weighting factors a ₀ , a _x and a ₂ , the sum of which a ₀ + ai + a ₂ = 1, in order to determine the filtered n-th data value y _n = a ₀ x _n + aχ _n - ₁ + a ₂ Xn- ₂ . The weighting factors can e.g. B. a ₀ = 0.25, ai = 0, 5 and a ₂ = 0.25, which means arithmetically only a shift of the bits and is therefore particularly suitable for digitized data values. The low-pass filter is one-dimensional because only the data occurring in a linear sequence are used one after the other. Thus, only the data determined one after the other in a line-by-line scan of the image are taken into account and not data which result from image points adjacent above and below the line in question.

The following is a more detailed description of an example of the fingerprint sensor with reference to FIGS. 1 and 2.

Figure 1 shows a schematic of the fingerprint sensor in an oblique view.

FIG. 2 shows a diagram of a rastered arrangement of sensor elements.

FIG. 1 shows an oblique top view of a semiconductor chip 1 with a fingerprint sensor integrated therein.

On the top there is a support surface for a finger, in or under which the sensor elements 2 are arranged, which in this example are simply formed by square conductor surfaces. A control circuit 3, with which the sensor elements, the data of which are to be recorded, can be addressed, can be integrated in the semiconductor chip 1, as in FIG. 1 by the area delimited by dashed lines is indicated. Electrical connections 4 are provided for controlling the fingerprint sensor by a micro-controller or the like and for outputting the digitized data.

This fingerprint sensor has another electronic circuit with which the captured and digitized data is compressed before being output. In the exemplary embodiment in FIG. 1, this further electronic circuit 5 is arranged in an area of the semiconductor chip 1 that is free of the sensor elements 2. The data reduction is therefore carried out "on the fly" directly in the hardware of the fingerprint sensor implemented in the semiconductor chip. A digital low-pass filter can also be part of this electronic circuit 5 or the control circuit 3. The data is preferably output serially in order to make do with as few electrical connections 4 as possible. A parallel interface is generally also suitable. In the example shown in FIG. 1, the number of connection contacts is chosen arbitrarily and can be different.

The data compression of an arrangement of pixels in a rectangular grid is preferably carried out row by row or column by column. It is e.g. B. from the data value of a first sensor element of a line. The data value of the subsequent sensor element is replaced by the difference between this data value and the data value of the first sensor element. The data value of the third sensor element of the line is replaced by the difference between this data value and the sum of the first data value of the line concerned and the difference which replaced the previous data value.

FIG. 2 shows a rastered arrangement of conductor surfaces of the sensor elements, the rows of which are each labeled with different capital letters and the columns of which are each labeled with different lower case letters. The one from that Gray values detected by sensor element Xy are referred to below as L (Xy). As an example, let us assume that the first sensor elements of the first line deliver the following gray values: L (Aa) = 5, L (Ab) = 20, L (Ac) = 21 etc. B. 5 bits per pixel are provided for coding, data values of the gray scale levels from -15 to +15 can be encoded. In this example, the gray value 5 of the first sensor element Aa is coded unchanged as data value 5 = 0101 (binary), a fifth bit not shown here indicating the sign (here: +). Instead of the gray value 20 of the second sensor element Ab, the difference to the data value of the first sensor element Aa, which here is 15 = 1111, is coded with 5 bits. Instead of the data value 20, the data value 15 is therefore coded, which is also the highest permissible data value. Instead of the gray value 21 of the third sensor element Ac, the difference between this gray value and the sum of the previously coded data values is coded, ie 21 - (5 + 15) = 1 = 0001.

This ensures that, with the exception of errors at individual pixels, the correct gray values also at the end of one

Section can be recovered from the coded data values without the occurrence of cumulative errors, as the example shows L (Aa) = 4, L (Ab) = 22, L (Ac) = 25, L (Ad) = 19. Aa 4 = 0 I 00, Ab 15 = I I I I (error) and Ac 25 - (4 + 15) = 6 = 0110 are encoded as data values. For Ab, instead of the correct data value 22 - 4 = 18 of the difference between the gray values of Ab and Aa, the wrong data value 15 is coded; in the subsequent formation of differences, this error is compensated for by always forming the difference from the sum of the previously coded differences. Consequently, that becomes

Gray value 25 of the sensor element Ac does not form the difference 3 to the gray value 22 of the sensor element Ab, but the difference 6 = 0110 to the sum of the first coded data value 4 = 0100 (to Aa) and the previously coded differences, in this case only the one Difference 15 = 1111 (to Ab). The sum of the coded data values up to and including the sensor element Ac then delivers the correct gray value again Ac: 4 + 15 + 6 = 25. Not the difference 19 - 25 = - 6 to the gray value 25 of Ac is formed for the sensor element Ad with the gray value 19, although in this case this also gives the correct result, but the difference to the sum of the data values already encoded: 19 - (4 + 15 + 6) = - 6.

It can therefore be assumed that, as a rule, with this type of deterministic difference coding, differences are formed up to the end of the line (sensor element Aj), which can be clearly converted into the correct gray values. The actual gray value of the first sensor element Ba is encoded again in the next line. The application of the difference coding in sections ensures that any errors that may occur permanently are completely eliminated in the planned steps of the calculation. The sections do not need to include entire lines of the arrangement of the sensor elements. The sections can be longer or shorter and optionally also have different lengths.

If the grid of sensor elements shown in FIG. 2 is scanned line by line from left to right and from top to bottom and the sections are selected to be eight sensor elements long, the sections end with the sensor elements Ah, Bf, Cd, Db, Dj, Eh, Ff, Gd, Hb, Hj, Ih, Jf and Jj. If the grid instead z. B. is scanned diagonally from top left to bottom right, starting with the sensor element Aj, and the sections each comprise five sensor elements, the sections end at the sensor elements Bi, Dj, Ξj, Ei, Dg, Bd, Gi, De, Ij , Ee, Jj, Fe, Ca, Hf, Fc, Ea,

Jf, Je, Ha and Ja. The first sensor elements of these sections are those whose data values after digitization have the maximum number of z. B. 8 bits per pixel and otherwise output unchanged. The simplest in a rectangular grid, however, is to subdivide into sections of entire lines, with in the example of FIG data values originating from the sensor elements of the first column a are encoded with the full number of bits.

The advantages of this method implemented in the fingerprint sensor are in particular the data reduction and the reduction in the bandwidth of the interface due to the predetermined fixed compression rate. A serial interface can be used instead of a parallel interface. Therefore, fewer electrical connections are required. The fingerprint sensor can be placed further away from the host system. The electricity consumption is reduced. The number of lines on the PCB board is reduced. The electronic circuits of the semiconductor chip, in particular the electronic circuit provided for data compression, can be implemented without a significant increase in chip area and power consumption. A synthesis approach (for example VHDL) is also suitable as a design methodology for the principle described.

LIST OF REFERENCE NUMBERS

1 semiconductor chip

2 sensor element

3 control circuit

4 electrical connection

5 electronic circuit

Claims

claims

1. Fingerprint sensor in a semiconductor chip (1) with an arrangement of sensor elements (2) for recording pixels, a control circuit (3) for detecting and digitizing values recorded by the sensor elements and outputs for outputting data values, characterized in that an electronic circuit (5) is integrated, which is provided for a reduction or compression of the data values, and the reduced or compressed data values are fed to the outputs.

2. Fingerprint sensor according to claim 1, in which the data values recorded and digitized by the sensor elements (2) are replaced in sections in a predetermined sequence with the exception of the data value of the first sensor element of a respective section by the difference between the data value of the respective sensor element and the Sum of the data value of the first sensor element in the relevant section and the differences by which the data values of the sensor elements following the first sensor element were replaced.

3. Fingerprint sensor according to claim 2, in which a row and column arrangement of the sensor elements

(2) is present, a line-by-line sequence of the sensor elements is specified, a section in which replacements are made is in each case a line and the digitized data value recorded and digitized by the first sensor element in each line is retained unchanged.

4. Fingerprint sensor according to claim 1 or 2, wherein an arrangement of the sensor elements (2) as a matrix, a strip or a single line is present.

5. Fingerprint sensor according to one of claims 1 to 4, in which the outputs are provided for outputting data values as a serial interface.

6. Fingerprint sensor according to one of claims 1 to 5, in which the sensor elements (2) comprise conductor areas and record a respective gray value of an associated pixel by a capacitive measurement.

7. Fingerprint sensor according to one of claims 1 to 6, in which the digitized data values pass through a low-pass filter integrated in the semiconductor chip (1) before the reduction or compression.

8. Method for operating a fingerprint sensor, in which values are acquired pixel by pixel and digitized into data values, characterized in that, before the data values are output to a circuit provided for evaluating the data values, the data values in sections in a predetermined sequence with the exception of the data value of the first pixel of each section are replaced by the difference between the data value of the respective pixel and the sum of the data value of the first pixel in the section concerned and the differences by which the data values of the pixels following the first pixel were replaced.

9. The method according to claim 8, in which a line-by-line sequence of the pixels is predetermined, a section in which replacements are made is one line at a time and the data value belonging to the first pixel of a line is retained unchanged.