JPH0990992A - Broad-band speech signal restoration method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain a broad-band speech signal from a narrow-band voice signal by efficiently using an existing telephone system. SOLUTION: This method consists of such processes as a 16kHz sampling processing 101 for 8kHz sampling narrow-band speech signal, a full-wave rectification processing 102 for up-sampled speech signal, a short-time Fourier analysis processing 103 for the full-wave rectification processed speech signal, a band-pass filtering processing 104 for low range of the complex spectra, a processing 106 to obtain a high range complex spectra from the complex spectra obtained by the processing 105, a STFT synthesis processing 109 to synthesize the complex spectra of the high and low ranges, and an addition processing 110 to sum the up-sampled speech signal and the STFT synthesized speech signal.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a wideband voice signal restoration method for restoring a wideband voice signal from a narrowband voice signal, and more specifically, a narrowband voice signal output by telephone voice or AM radio. The signal is sent to the audio set or F
The present invention relates to a wideband audio signal restoration method for improving the quality of a wideband audio signal output by an M radio or the like.

[0002]

2. Description of the Related Art Conventionally, narrow band voice signals have been used in telephone systems. The signal spectrum band of the narrow band voice signal that can be transmitted by the existing telephone system is about 3
It is from 00 Hz to 3.4 kHz. The purpose of conventional voice coding techniques is to maintain the quality of voice in the telephone band and to minimize the amount of transmission parameters. Therefore, with the conventional audio encoding technique, it is possible to reproduce the input voice, but it is impossible to obtain a voice that exceeds the quality of the input voice.

[0003]

By the way, since the quality of the sound used in daily life is improved by the recent development of the acoustic technology and the development of the digital processing, the sound quality of the current telephone band is high. The current situation is that it is not possible to meet the demands of telephone users. As a method of solving such a request, the existing telephone system is destroyed,
Reconstruction of a telephone system capable of transmitting a wideband voice signal is considered, but this method is not only economically burdensome, but also considered to require considerable time for reconstruction. It is poor in feasibility. Further, as means for solving the above-mentioned problems, proposals of JP-A 06-118995, JP-A 07-56599, and Japanese Patent Application No. 06-209622 have already been made, but they have a complicated structure. I'm not satisfied. The present invention has been made under such a background, and effectively utilizes an existing telephone system to obtain a wideband voice signal and to transmit a wideband signal and an existing telephone system. It is an object of the present invention to provide a wideband voice signal restoring method capable of obtaining a wideband voice signal regardless of the combination of both telephone systems even in the situation where the above (narrowband) telephone system coexists.

[0004]

According to a first aspect of the present invention, a narrowband speech signal is nonlinearly processed to generate a wideband signal having a wider band than the narrowband speech signal, and the narrowband speech signal is included in the narrowband speech signal. A low-frequency signal or a high-frequency signal that is not present is extracted from the wide-band signal, and the narrow-band audio signal and the low-frequency signal or the high-frequency signal are combined.

According to a second aspect of the present invention, a narrowband speech signal is nonlinearly processed to generate a wideband signal having a wider band than the narrowband speech signal, and a lowband signal not included in the narrowband speech signal and Extracting a high frequency signal from the wide band signal,
It is characterized in that the narrow band voice signal is combined with the low band signal and the high band signal.

According to a third aspect of the present invention, a narrow band speech signal is spectrally analyzed, a signal having a low band spectral component obtained as a high band spectral component is restored, and this restored signal is used as the narrow band signal. It is characterized in that it is synthesized with an audio signal.

[0007]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a flowchart showing a procedure of a wideband speech signal restoration method according to an embodiment of the present invention. The process described below is performed by a computer.

In this figure, in process 101, 8 kHz
The z-sampled narrowband speech signal is upsampled from an 8 kHz sampling signal to a 16 k sampling speech signal, whereby a wider-band speech signal than the narrowband speech signal is obtained.

In the process 102, the up-sampled voice signal is full-wave rectified (non-linear process) to obtain a wide band voice signal having a high correlation with the narrow band voice signal, including a harmonic component of the fundamental frequency of the voice.

In the process 103, a fixed time (10 to 20 m
sec), the full-wave rectified audio signal is S
TFT (short-time Fourier) analysis is performed, and the following complex spectrum Sj is obtained. Sj = Xj + iYj In the complex spectrum Sj, the subscript j represents the frequency band number, Xj represents the real part, and Yj represents the imaginary part.

In process 104, bandpass filtering in the frequency domain is performed to obtain the complex spectrum S.
Of j, a complex spectral component of 300 Hz or less that does not exist in the narrow band audio signal is extracted. On the other hand, 60Hz
The following complex spectral components are removed. That is, in the process 104, the complex spectrum of the low-frequency signal (60 to 300 Hz) that is not included in the narrowband audio signal is obtained. It should be noted that the signal of 60 Hz or less is generally not included in the audio signal, and is often audible to the ear, so that removal of the signal is important. Processing 10
In 7, the low-pass band-pass filtered complex spectrum is multiplied by a constant scale factor, and the gain is adjusted accordingly.

In parallel with the above processes 102 to 104, in process 105, every time a fixed time (10 to 20 msec) elapses, the upsampled audio signal is subjected to STFT analysis, and the following complex spectrum Sj 'is obtained. Sj '= Xj' + iYj '

In step 106, the complex spectrum of the low frequency band is copied as the complex spectrum of the high frequency band. For example, the 10th to 40th complex spectra are used as the 100th to 130th complex spectra. Also, all 0s are substituted for the complex spectrum of the other part. As a result, a high frequency complex spectrum that is not included in the narrow band audio signal is obtained. In the process 108, the complex spectrum in the high frequency band is multiplied by a constant scaling factor, and the gain is adjusted accordingly.

In process 109, the low frequency complex spectrum obtained in process 107 and the high frequency complex spectrum obtained in process 108 are subjected to STFT synthesis. As a result, low-frequency and high-frequency audio signals not included in the narrowband audio signal can be obtained.

In process 110, the voice signal up-sampled in process 101 and the voice signal synthesized with STFT in process 109 are added. As a result, a pseudo wideband audio signal having low and high frequencies not included in the narrowband audio signal can be obtained.

Although the embodiment of the present invention has been described in detail above with reference to the drawings, the specific structure is not limited to this embodiment, and design changes and the like within the scope not departing from the gist of the present invention. Even if it exists, it is included in the present invention. For example, in the wideband audio signal restoration method according to the above-described embodiment,
An example in which full-wave rectification (process 102) is performed as the non-linear process has been described, but other non-linear process may be performed instead.

In the wideband speech signal restoration method according to the above-described embodiment, an example of performing filtering processing (processing 103, 104) in the frequency domain has been described.
The same process may be performed by FIR filtering or IIR filtering. Furthermore, in the wideband speech signal restoration method according to the above-described embodiment, the example in which the high frequency complex spectrum is generated by the processes 105 and 106 has been described, but the processes 102 to 104 may be performed instead of these processes. . In this case, it is necessary to perform high-pass filtering instead of band-pass filtering in process 104.

[0018]

As described above, according to the present invention, in a method for restoring a wideband speech signal from a narrowband speech signal, the characteristics of the speech signal which do not exist in the narrowband speech signal are efficiently processed by the non-linear processing. Can be restored. Moreover, it is possible to handle wideband audio signals at low cost by only modifying a part of the existing system.

[Brief description of drawings]

FIG. 1 is a flowchart showing a procedure of a wideband speech signal restoration method according to an embodiment of the present invention.

[Explanation of symbols]

 101 Upsampling process 102 Full-wave rectification process 103, 105 STFT analysis process 104 Bandpass filtering process 106 High frequency generation process 107, 108 Multiplication process 109 STFT synthesis process 110 Addition process

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI Technical display location H04M 1/00 H04M 1/00 H

Claims (3)

[Claims] 1. A non-linear processing of a narrow band speech signal to generate a wide band signal having a wider band than the narrow band speech signal, and a low band signal or a high band signal not included in the narrow band speech signal is converted into the wide band signal. A wideband speech signal restoration method comprising: synthesizing the narrowband speech signal with the lowband signal or the highband signal. 2. A non-linear processing of a narrow band speech signal is performed to generate a wide band signal having a wider band than the narrow band speech signal, and a low band signal and a high band signal which are not included in the narrow band speech signal are the wide band signal. A wideband speech signal restoration method comprising: synthesizing the narrowband speech signal with the lowband signal and the highband signal. 3. A spectrum analysis of a narrow band speech signal is performed, a signal having a low band spectrum component obtained as a high band spectrum component is restored, and the restored signal is combined with the narrow band speech signal. Characteristic wideband speech signal restoration method.