Signal detector with coefficient swapping transform unit

1. Field of the Invention

The present invention relates generally to the detection of signals, such as dual tone multi-frequency (DTMF) coded signals, and more particularly, to reducing the complexity of a signal detector by swapping coefficients in the transform unit used to determine signal energies at targeted frequencies.

2. Description of the Related Art

DTMF is the signaling standard for the Touch Tone system. Two tones are combined to define a DTMF digit. Table 1 shows the matrix of frequencies used to encode the 16 DTMF symbols on a typical 4×4 matrix telephone keypad. Each DTMF tone consists of two sine waves. One sine wave is from the low-frequency band (row frequency), and the other frequency is from the high-frequency band (column frequency).

Various methods have been used for DTMF generation and detection. Previously, analog circuitry had been used to generate and decode DTMF digits. More recently, the functions have been implemented using digital signal processors (DSP) to improve the accuracy, precision, stability, reprogrammability, and cast. A popular DSP algorithm for DTMF detection is the Goertzel algorithm. One such previous method is described in U.S. Pat. No. 5,644,634, issued to Zheng-yi Xie et al., entitled “System and Method for Dual Tone Multifrequency Detection Using Variable Frame Widths,” the specification of which is hereby incorporated by reference in its entirety.

In general, a DTMF detector examines the line or communication channel for the presence of two sinusoids using dedicated frequency domain algorithms, including modified Goertzel algorithms, DFT/FFTs, auto-correlation, zero crossing counting, and narrow band filter-based methods, among others.

A common technique for determining energy values in the frequency domain is performed by transforming the time domain sampling data to frequency domain data using a variant of the discrete Fourier transform (DFT) called the Goertzel algorithm. The transfer function for the Goertzel algorithm is: H k = 1 -  j     2     π     k N 1 - c · z - 1 + z - 2 , ( 1 )

where c is the Goertzel coefficient, c=2 cos ( 2     π     k N )

and N is the number of samples in a frame. The value of k is defined by: k = N F s · F tone , ( 2 )

where F s is the sampling frequency (i.e., 8 kHz in typical telephone systems), F tone is the frequency nearest the DTMF frequency that keeps the value of k to an integer. The energy of each DTMF frequency is calculated by the feed forward and feedback phase. The feedback phase is calculated as:

Q ( n )= c·Q ( n− 1)− Q ( n− 2)+ x ( n ),  (3)

where Q(−1)=Q(−2)=0, n=0, 1, 2, . . . , N−1. Q(n−1) and Q(n−2) are feedback storage elements for a frequency point, and x(n) is the current input sample value. Q(n) is referred to as the delay value.

The energy in each DTMF frequency bin is calculated in the feedforward phase (i.e., n=N) by the equation:

E ( dtmf )= Q 2 ( n− 1)+ Q 2 ( n− 2)− Q ( n− 1)· Q ( n− 2)· c.   (4)

After the energy of each DTMF frequency is calculated, the largest magnitude in each frequency band is compared to a threshold. If the energy in one of the bins exceeds the threshold, a DTMF tone is present.

In accordance with industry standards, a DTMF detector should detect a DTMF tone have less than 1.5% frequency distortion and reject a tone having more than 3.5% distortion. A commonly used twist test detects an energy variation due to an amplitude variation. The allowable energy difference between row and column frequencies for a given detection is between −8 dB and 4 dB (i.e., amount of twist). The twist test is not effective in determining energy variations due to frequency distortion. For example, consider the case where the row frequency of Digit “1” varies by 3%. The energy of the row frequency drops by about 1 dB, but the twist test limits are still met. Current DTMF detectors have difficulty discriminating between energy variations due to frequency variation and those caused by amplitude variation, and accordingly, the frequency selectivity of the detector is reduced.

The present invention is directed to overcoming, or at least reducing the effects of, one or more of the problems set forth above.

One aspect of the present invention is seen in a signal detector including a transform unit and a post processing test unit. The transform unit is adapted to a transform unit adapted to receive at least first and second frames of input samples, determine a first plurality of energy values for the first frame based on a first set of transform coefficients, and determine a second plurality of energy values for the second frame based on a second set of transform coefficients. The post processing test unit is adapted to identify at least one target frequency based on the first plurality of energy values. The second set of transform coefficients is based on the target frequency, and the post processing unit is adapted to determine the presence of a first signal in the input samples based on the second plurality of energy values.

Another aspect of the present invention is seen in a method for detecting a signal. A first frame of input samples is received. A first set of transform coefficients is determined based on a first plurality of frequencies. A first plurality of energy values is determined based on the first set of transform coefficients. A target frequency is determined from the first plurality of frequencies based on the first plurality of energy values. A second set of transform coefficients is determined based on the target frequency. A second frame of input samples is received. A second plurality of energy values is determined based on the second set of transform coefficients. The presence of the signal in the input samples is determined based on the second plurality of energy values.

Other objects and advantages of the invention will become apparent upon reading the following detailed description and upon reference to the drawings in which:

FIG. 1 is a block diagram of a signal detector in accordance with the present invention;

FIG. 2 is a functional block diagram of the signal detector of FIG. 1; and

FIG. 3 is a flow chart illustrating a coefficient swapping technique implemented by the signal detector of FIG. 1 .

While the invention is susceptible to various modifications and alternative forms, specific embodiments thereof have been shown by way of example in the drawings and are herein described in detail. It should be understood, however, that the description herein of specific embodiments is not intended to limit the invention to the particular forms disclosed, but on the contrary, the intention is to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the invention as defined by the appended claims.

Illustrative embodiments of the invention are described below. In the interest of clarity, not all features of an actual implementation are described in this specification. It will of course be appreciated that in the development of any such actual embodiment, numerous implementation-specific decisions must be made to achieve the developers' specific goals, such as compliance with system-related and business-related constraints, which will vary from one implementation to another. Moreover, it will be appreciated that such a development effort might be complex and time-consuming, but would nevertheless be a routine undertaking for those of ordinary skill in the art having the benefit of this disclosure.

Referring to FIG. 1, a block diagram of a signal detector 100 is provided. The signal detector 100 includes a receiver 102 , a digital signal processor 104 , and a memory 106 . The receiver 102 receives an analog signal and converts the analog signal into digital format. The receiver 102 samples the analog signal and utilizes pulse code modulation (PCM) or other suitable techniques to produce corresponding digital data. The receiver 102 may include linear analog to digital (A/D) converters (not shown). In the illustrated embodiment, the receiver 102 samples at a frequency of 8 kHz, a common sampling frequency used in telephone systems.

It is noted that the received signal may be compressed or companded, and thus digital data produced by the A/D converter (not shown) in the receiver 102 may be companded (i.e., may comprise logarithmically compressed digital data). As is well-known in the art, companding refers to logarithmically compressing a signal at the source and expanding the signal at the destination to obtain a high end-to-end dynamic range while reducing dynamic range requirements within the communication channel. If the signal is companded, the receiver 102 logarithmically expands the data to a linear format. It is also contemplated that the signal detector 100 may receive digital signals (i.e., linear or logarithmic) directly, and thus a receiver 102 is not required in the signal detector 100 for analog to digital conversion.

The frequency domain energy values calculated for the DTMF frequencies are used to determine the presence of a DTMF signal. A variety of tests known in the art may be performed to determine the presence of a valid DMTF signal and to reject false detections caused by imitation signals, such as noise, voice, or music. The series of tests performed on the energy values are referred to as post-processing tests. In the illustrated embodiment, the generation of the energy values and the post processing tests are performed by the digital signal processor 104 .

The memory 106 may be a read-only memory, a random access memory, or a combination thereof. The memory 106 may be employed to store program instructions for execution by the digital signal processor 104 . Also, the results of calculations completed by the digital signal processor 104 may be stored in the memory 106 . It is also contemplated that a microprocessor (not shown) or other type of general purpose processor (not shown) may be used to perform any or all of the functions of the receiver 102 and/or the digital signal processor 104 . In the illustrated embodiment, the digital signal processor is an ADSP-2181, sold by Analog Devices, Inc.

A functional block diagram of the signal detector 100 is provided in FIG. 2 . The signal detector 100 includes the receiver 102 adapted to sample the input signal, and a Goertzel Discrete Fourier Transform (GDFT) unit 200 adapted to transform the time domain samples of the receiver 102 to frequency domain energy values using the equations described above. The GDFT unit 200 is adapted to determine energy values for the DTMF frequencies, their second harmonics, and specified cutoff frequencies for testing the frequency distortion. The GDFT unit 200 does not calculate all of these values for every frame. As described in greater detail below in reference to FIG. 3, the signal detector 100 employs a coefficient swapping technique to determine the energies of the eight DTMF frequencies for a first frame, and in response to detection a DTMT signal, determines the energy of the identified tones, their second-harmonics, and their cutoff frequencies.

For purposes of illustration, exemplary post processing tests are described. It is contemplated that only a portion of these tests may be employed by the signal detector 100 , or other additional tests may be included. To perform the post processing tests the signal detector 100 includes a gain unit 202 adapted to scale the output of the GDFT unit 200 based on any asymmetry in frame widths between the frequencies. A signal detector having frame widths optimized for the various DTMF frequencies is described in U.S. patent application Ser. No. 09/196,267, entitled “SIGNAL DETECTOR WITH DURATION BASED FRAME WIDTH,” and incorporated herein by reference in its entirety.

The signal detector 100 also includes an energy level test unit 204 adapted to compare the largest row and column DTMF energy values to a minimum threshold, and a twist test unit 206 adapted to determine if the power level difference between the largest row and column energy values is within a limit. A frequency offset test unit 208 is adapted to determine if the frequency offset of each DTMF frequency is within a limit, and a talk-off test unit 210 is adapted to compare the second harmonics of the DTMF frequencies to a harmonic threshold to determine if a DTMF detection might have been simulated by speech. The harmonic threshold may be static (i.e., a fixed value) or dynamic (i.e., based on the DTMF energy).

A guard time check unit 212 is adapted to determine if the DTMF tone has an acceptable duration (i.e., greater than 20 ms). In the illustrated embodiment, the frame size is about 106 samples (i.e., 13.25 ms). If at least two consecutive detections for a particular digit are identified, the guard time is inherently met. Functions and algorithms to implement the post processing tests described above are known in the art and, for clarity, they are not further described herein unless necessary to illustrate the application of the present invention. Also, it is contemplated that the tests illustrated in FIG. 2 may be performed in an order different from that shown, depending on the specific implementation. Collectively, the gain unit 202 , energy level test unit 204 , twist test unit 206 , frequency offset test unit 208 , talk-off test unit 210 , and guard time check unit 212 may be referred to as a post processing unit 214 .

The operation of the frequency offset test unit 208 is now described in greater detail. To identify frequency distortion, the frequency offset test unit 208 evaluates the energy at the DTMF frequency and the values at the cutoff frequencies. If the cutoff frequency energy value times a frequency offset threshold exceeds the DTMF frequency energy value, the detection is abandoned. The frequency offset test is defined by the equation:

E DTMF >T FO ·E CO   (5)

In the illustrated embodiment, the cutoff frequencies and frequency offset threshold are set such that a detection is abandoned if the frequency offset is greater than about ±2.5%. In the illustrated embodiment, the frequency offset threshold is about 10 dB. It is contemplated that other cutoff frequencies and frequency offset thresholds may be used depending on the degree of frequency selectivity desired and the specific cutoff frequencies. Table 2 illustrates exemplary cutoff frequencies used for the frequency offset determination in the illustrated embodiment.

If the energies at the cutoff frequencies were to be determined for each frame, the computational resources and memory required to support the GDFT unit 200 would double (i.e., 16 Cut-off energies+existing 16 DTMF and harmonic energies). To avoid this resource burden, the signal detector 100 employs a coefficient swapping technique.

Turning to FIG. 3, a flow chart of the coefficient swapping technique is illustrated. The GDFT unit 200 is loaded with the coefficients necessary to calculate the energy values at the eight DTMF frequencies (DTMF coefficients) in block 400 . The GDFT unit 200 determines the DTMF frequency energy values for a first frame in block 405 . An energy level and twist test are performed in block 410 by the energy level test unit 204 and the twist test unit 206 , respectively. If no DTMF digit is identified in block 415 based on the energy level and twist tests, the GDFT unit 200 calculates the DTMF frequency values for the next frame in block 405 , and the process continues until a valid digit is identified.

If a valid DTMF digit is detected in block 415 , the GDFT unit 200 is loaded with coefficients for calculating the energies of the targeted DTMF frequencies for the identified digit (i.e., one row frequency and one column frequency), the second harmonics of the targeted frequencies, and the cutoff frequencies for the targeted frequencies (target coefficients) in block 420 . In block 425 , the energy values of the targeted DTMF, harmonic, and cutoff frequencies for a second frame are determined. In the illustrated embodiment, the sample size for the DTMF frequencies is optimized for each frequency. The sample size for the second harmonics and the cutoff frequencies is about 106 samples, however, it is contemplated that these frequencies may also be optimized.

In block 430 , the energy level and twist tests are performed for the targeted DTMF frequencies. The frequency offset test is performed by the frequency offset test unit 208 as described above to determine if the frequency distortion is within allowable tolerances. The talk-off test unit 210 determines if the second harmonics of the targeted frequencies are below a specified threshold, and the guard time check unit 212 verifies that the tone is of sufficient duration (e.g., 20 ms).

If all of the post processing tests performed in block 430 pass, a valid DTMF digit is identified in block 435 , and the DTMF digit is outputted in block 440 . If the post processing tests of block 430 do not all pass, the GDFT unit 200 is loaded with the DTMF coefficients in block 405 and detection continues on subsequent frames.

During or after completion of the post processing tests described above, it is contemplated that the digital signal processor 104 may set a flag (not shown) in the memory 106 to indicate the detection or non-detection of a DTMF signal.

If a valid DTMF signal for a digit other than the targeted digit were to be received during the second frame, it might be missed due to the targeted coefficients. This situation might occur if a false detection is generated for the first frame. In the illustrated embodiment, the frame size is about 106 samples, or 13.25 ms. Even if the true DTMF signal were not detected until the third frame (i.e., first frame—false detection, second frame—missed due to targeted coefficients, third frame—valid DTMF signal), the valid DTMF signal would still be detected. Per industry standards, a DTMF signal should have a duration of at least 40 ms. Accordingly, the valid DTMF signal would be detected in frames three and four by the method described above.

Swapping the DTMF coefficients and the targeted coefficients during the detection process reduces the complexity of the GDFT unit 200 , and accordingly, the resources needed to support it, by one half as compared to a traditional DTMF detector (not shown). Only eight energy values are computed for each frame as opposed to the 16 computed for the traditional DTMF detector (not shown). Also, if the traditional DTMF detector (not shown) were to be augmented to add the frequency offset test, 32 energy calculations would be required.

Although the coefficient swapping technique is described as it may be used to generate cutoff frequency energy values without increasing complexity, it may also be applied to a signal detector 100 for calculating standard DTMF energy values. For example, in the first frame the energy values for all of the DTMF frequencies are calculated, and a target digit is identified. On the second frame, only the target frequencies and the harmonics are calculated. Second and third harmonic energies may also be calculated in lieu of the frequency offset energies. Accordingly, the GDFT unit 200 need only calculate a maximum of eight energy values for any given frame. As a result, the processing requirements of the digital signal processor 104 could be reduced, thus reducing the cost of the signal detector 100 .

The particular embodiments disclosed above are illustrative only, as the invention may be modified and practiced in different but equivalent manners apparent to those skilled in the art having the benefit of the teachings herein. Furthermore, no limitations are intended to the details of construction or design herein shown, other than as described in the claims below. It is therefore evident that the particular embodiments disclosed above may be altered or modified and all such variations are considered within the scope and spirit of the invention. Accordingly, the protection sought herein is as set forth in the claims below.