Telephonic communications – Supervisory or control line signaling – Signal receiver
Reexamination Certificate
1999-10-06
2004-01-06
Tieu, Binh (Department: 2643)
Telephonic communications
Supervisory or control line signaling
Signal receiver
C379S283000
Reexamination Certificate
active
06674855
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to detection of multifrequency (MF) communication signals such as dual-tone multifrequency (DTMF) signals.
2. Description of the Related Art
Multifrequency (MF) signaling allows control and data signals to be transmitted over conventional communication channels that may be used concurrently to carry other information, such as voice and data. A MF signal comprises two or more simultaneous (and relatively pure) tones that together represent a digit, character, or other information according to an agreed format. The communication channel can be provided by land lines (wire, cable, fiber optic, etc.), wireless signaling such as microwave and satellite links, and so forth.
Telephone networks use Dual-Tone Multi-Frequency (DTMF) signaling to transmit dialed digits between telephone sets and a central office or switch. Typically DTMF techniques are also used for voice-mail user interfaces, Interactive Voice Response (IVR) systems, and similar interactive control systems. DTMF signaling is a particular case of MF signaling that will be discussed herein in detail. However, the same issues of concern for DTMF signaling also arise in MF signaling generally.
A DTMF signal is a combination of two tones having well-defined frequencies typically selected from a standard pair of disjoint frequency band. For example, the current ITU-T standard (set forth in the ITU-T Q.24 recommendation) establishes a high band tone set of four tones with nominal frequencies at 1209, 1336, 1477, and 1633 hertz (Hz), respectively. The low band consists of four tones with nominal frequencies at 697, 770, 852, and 941 Hz. Of course, alternate MF standards could use different nominal frequencies and could use more or fewer tones in either or both bands.
A MF signal is detected using a MF detector. Such a product must be highly reliable to ensure proper detection of information represented in a valid MF signal while rejecting detection of non-MF signals that may also be received on the communication channel. A MF detector, such as a DTMF detector, must be able to determine (reliably) that a true MF signal (such as a dialed “7”) has been received. Determination of whether a received signal is an MF signal will be “digit detection.” The detector must also be able to identify (reliably) the tone components representing the dialed digit, character, or other information.
High quality MF detectors must be designed for robustness in conditions of substantial signal impairments. Good signal detection performance means the MF detector must be capable of reliable digit detection in the presence of voice, non-MF data, random noise, and so forth (collectively “noise”). This entails a requirement that the detector have both high sensitivity and high selectivity. “Sensitivity” refers to the degree of assurance that the detector will correctly detect a DTMF signal when such a signal is received. “Selectivity” refers to the degree of assurance that the detector will reject a received signal when the signal is actually not a valid MF signal.
High selectivity is reflected in a low “talk-off” rate, which is defined as a low rate at which non-MF signals are incorrectly identified as valid MF signals. Talk-offs typically arise from noise (such as voice signals, non-MF data signals, random noise, etc.). High sensitivity is reflected in a low “talk-down” rate, i.e., a low rate at which valid MF signals are incorrectly classified as non-MF signals. Talk-downs may occur when true MF signals are transmitted, but one or more of their constituent tones are shifted in frequency from their nominal frequencies. A talk-down also can occur when a constituent tone in a valid MF is misshapen by the communication link and therefore contains frequency constituents other than the nominal frequency for that tone. A high quality detector must have both a low talk-off rate and a low talk-down rate.
Noise and distortion can also degrade the detector's ability to identify correctly the digit, character, or other information represented by a valid MF signal. For example, a high quality DTMF detector must be able to identify correctly the two nominal tones of a DTMF signal. The tone identification must be highly reliable, even when the tones are subjected to frequency offsets and are superimposed with noise.
Most commercial DTMF detectors perform spectral computations, particularly based on modified Goertzel (MG) algorithms, to analyze incoming signals. A spectral decomposition of the signal is thereby generated, with a calculated energy for each of several nominal frequencies. Such detectors differ mainly in the way the results of the energy calculations are used for signal detection, i.e., to distinguish between true DTMF signals and other signals or noise.
A typical Goertzel-based detector will have a digital filter providing, for example, eight passbands centered at the eight nominal frequencies of the conventional DTMF standard. This basic structure constrains the computational and signal accumulation requirements of the detector within reasonable limits. On the other hand, such detectors are prone to false detections (talk-offs) when a non-DTMF signal has high and low band energy peaks close to center frequencies for the filter. Most MG filters therefore are unable to reject a signal (i.e., correctly detect the signal as a non-DTMF signal) if the signal has energy peaks within 3% of nominal DTMF frequencies.
To avoid this trade-off between accumulation time and frequency resolution, it has been proposed to estimate the frequencies of the incoming signal directly. See David L. Beard,
DSP Implementation of a High Performance DTMF Decoder
, ICSPAT-98 Proceedings 448-452 (1998). A frequency estimator was constructed by opening the feedback path of a Goertzel filter, thereby adapting the structure to measure a frequency, instead of measuring energy at a specified frequency.
This frequency estimation approach was able to satisfy some performance standards by including a running statistical analysis to determine a mean and standard deviation of the frequency estimates. Separate statistics thresholds were used for digit detection and in-digit stability. Further, effective performance involved selection of thresholds and parameters specific to a desired performance standard. With careful tailoring of the statistics thresholds, this detector had good talk-off performance in the presence of frequency offsets.
Detectors based on MG filters are also prone to false rejections (talk downs) when the signal “twist” is too high. “Twist” is defined as the ratio of highest energy in the high band to the highest energy in the low band (expressed in dB). In a typical case, such a MF detector may be configured to reject signals having twist less than −4 dB or greater than 4 dB. If the communication channel induces significant frequency-dependent attenuation, then the detector will have a high rate of false rejections.
U.S. Pat. No. 5,442,696, issued Aug. 15, 1995, to Lindberg et al., describes an alternative to Goertzel algorithm filters in a DTMF detector. An incoming signal is sliced into successive frames, and each frame is duplicated in several copies. For each frame, the copies of the frame are shaped with different shaping sequences (“tapers”) such as discrete prolate spheroidal sequences.
The detector then Fourier-transforms the shaped copies and, like the MG-based detectors, combines the results to form an estimate of the signal's spectral distribution of energy. An inner product is then formed between the spectral estimate and each of several model energy distributions, and the inner product is compared with a threshold. Signal detection is then performed based on the inner products, if any, that pass the threshold test. Like MG filtering, therefore, this approach evaluates an incoming signal against discrete nominal frequencies and requires substantial computation and data accumulation for each frequency.
Canadian Patent No. 2,094,412 discloses a detector using an a
Karelic Yair
Shoham Tamar
Wexler Yosef
Comverse Ltd.
Tieu Binh
LandOfFree
High performance multifrequency signal detection does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with High performance multifrequency signal detection, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and High performance multifrequency signal detection will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3230665