Vehicle-mounted voice positioning method and system

Abstract

The invention discloses a vehicle-mounted voice positioning method and system, and relates to the technical field of automobile voice positioning. The method specifically comprises the following steps: defining the position of a sound acquisition device; acquiring audio data acquired by each sound acquisition device, and preprocessing each audio data to generate corresponding to-be-positioned data; each piece of to-be-positioned data is subjected to normalization processing, all normalization processing results are converted into positive values and then summed to generate a total audio energy value, the total audio energy value and the sampling number are subjected to quotient, average energy values are generated, and the sound acquisition device with the maximum average energy value in the sound acquisition devices is acquired and defined as a sound source position. The objective of the invention is to reduce the calculation complexity of sound source localization on the premise of ensuring the sound source localization precision.

Description

Technical Field

[0001] The present invention relates to the technical field of automobile voice positioning, and in particular to a vehicle-mounted voice positioning method and system. Background Art

[0002] At present, with the rapid development of the automobile industry and the improvement of intelligence, the in-vehicle voice control system is becoming an important part of modern automobiles. The in-vehicle voice control system can not only improve the convenience and safety of driving, but also provide more intelligent and personalized services for drivers and passengers. In the in-vehicle voice control system, speech recognition and sound source localization are two key technologies. Among them, the sound source localization technology can help the system accurately identify the voice commands of the driver or passenger, thereby improving the accuracy of the system response.

[0003] At present, mainstream in-vehicle voice localization systems usually use a multi-microphone array combined with a sound source localization algorithm. The multi-microphone array collects audio signals in the car by arranging multiple microphones, and uses a sound source localization algorithm to determine the location of the sound source. The advantage of this solution is that it can improve the accuracy of speech recognition and anti-interference ability through spatial information.

[0004] However, audio signal processing of multi-microphone arrays requires complex signal processing operations, including signal preprocessing, feature extraction, and sound source localization algorithms. These operations require high-performance processors and a large amount of memory resources to ensure that the system can complete the calculation in a short time and give accurate positioning results. In order to improve the accuracy of sound source localization, high-precision delay estimation and beamforming technology are required, which further increases the computational complexity of the system.

[0005] In order to improve the response speed of the system, some computing tasks are handed over to external servers for processing. However, this method relies on the stability of the network. Once the network fluctuates or delays, it will affect the system's response speed and user experience.

[0006] Therefore, how to reduce the computational complexity of sound source localization while ensuring the accuracy of sound source localization has become a technical problem that needs to be solved urgently. Summary of the invention

[0007] The main purpose of the present invention is to provide a vehicle-mounted voice localization method and system, aiming to reduce the computational complexity of sound source localization while ensuring the accuracy of sound source localization.

[0008] In order to achieve the above object, the present invention proposes a vehicle-mounted voice positioning method, comprising the following steps: Define the location of the sound collection device; Acquire the audio data collected by each sound collection device, pre-process each audio data, and generate corresponding data to be located; Each data to be located is normalized, all normalized results are converted into positive values ​​and then summed to generate a total audio energy value, the total audio energy value is divided by the sampling number to generate an average energy value, and the sound collection device with the largest average energy value among the sound collection devices is obtained, which is defined as the sound source position.

[0009] In one embodiment of the present application, the format of the audio data is RAW, the number of channels is 2 or 1, and the sampling frequency is 44100 Hz.

[0010] In one embodiment of the present application, preprocessing each audio data includes the following steps: Recording the ambient noise of the sound collection device; Performing Fourier transform on the audio data and the ambient noise to obtain an audio data spectrum and an ambient noise spectrum; Calculate and set the coefficient between the audio data spectrum and the ambient noise spectrum; According to the coefficient between the audio data spectrum and the ambient noise spectrum, the ambient noise spectrum is subtracted from the audio data spectrum to generate clean audio data.

[0011] In one embodiment of the present application, preprocessing each audio data further includes the following steps: A linear convolution operation is performed on the clean audio data to generate a smooth audio signal to reduce noise in the clean audio data.

[0012] In one embodiment of the present application, preprocessing each audio data further includes the following steps: Set the gain factor; The smooth audio signal is gain-adjusted according to the gain factor to increase the signal strength of the smooth audio signal.

[0013] In one embodiment of the present application, preprocessing each audio data further includes the following steps: A noise vector having the same length as the smoothed audio signal is generated, the noise vector is fused with the smoothed audio signal having completed gain, and the signal-to-noise ratio of the fused audio model is optimized.

[0014] The present application discloses a vehicle-mounted voice positioning system, comprising: A plurality of sound collecting devices for collecting audio data; and The processing unit processes the collected audio data unit using the vehicle-mounted voice positioning method as described above to obtain the sound source position corresponding to the audio data.

[0015] In one embodiment of the present application, the number of the sound collecting devices is four, which are evenly distributed on one side of the car seat.

[0016] In one embodiment of the present application, the sound collection device is a passive microphone.

[0017] By adopting the above technical solution, the sound source position in the car can be accurately located through the collaborative work of multiple sound collection devices. It can process audio data in real time and quickly locate the sound source position to meet the real-time needs in the vehicle environment. Its data processing process is simple, and the calculation complexity of sound source positioning is reduced while ensuring the accuracy of sound source positioning. It is easy to implement in the actual system and does not require complex hardware equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The present invention is described in detail below with reference to specific embodiments and accompanying drawings, wherein: Figure 1 It is a schematic structural diagram of the first embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present invention clearer, the present invention is described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the following specific embodiments are only used to explain the present invention and do not constitute a limitation to the present invention.

[0020] like Figure 1 As shown, in order to achieve the above purpose, the present invention proposes a vehicle-mounted voice positioning method, comprising the following steps: Define the location of the sound collection device; Acquire the audio data collected by each sound collection device, pre-process each audio data, and generate corresponding data to be located; Each data to be located is normalized, all normalized results are converted into positive values ​​and then summed to generate a total audio energy value, the total audio energy value is divided by the sampling number to generate an average energy value, and the sound collection device with the largest average energy value among the sound collection devices is obtained, which is defined as the sound source position.

[0021] Specifically, in the vehicle environment, multiple sound collection devices (such as microphones) are installed, and the positions of these sound collection devices need to be defined and fixed in advance. Determine the areas in the vehicle that need to be monitored, such as the driver's seat, passenger seat, back seat, etc. Select the corresponding sound collection device installation location based on the space in the vehicle and the area that needs to be monitored. Fix the sound collection device to ensure that its position is stable and unchanged to facilitate the consistency of subsequent data.

[0022] Each sound collection device will collect surrounding audio data in real time. These audio data need to be preprocessed to remove noise and interference to ensure the accuracy of subsequent positioning calculations. Start the sound collection device and start collecting audio data in real time. Preprocess the collected audio data, including but not limited to denoising, filtering, gain adjustment, etc. Save the preprocessed audio data as data to be positioned.

[0023] Normalize each data to be located and convert the sampled values ​​to the range of [-1, 1]. Ensure the consistency of the normalization process to avoid errors caused by different processing methods. Square each sampled value after normalization to ensure that all values ​​are positive. Sum all squared sampled values ​​to obtain the total audio energy value. Count the number of samples collected by each sound collection device. Divide the total audio energy value by the number of samples to obtain the average energy value. Compare the average energy values ​​of all sound collection devices. Find the sound collection device with the largest average energy value. Define the position of the sound collection device as the sound source position.

[0024] By adopting the above technical solution, the sound source position in the car can be accurately located through the collaborative work of multiple sound collection devices. It can process audio data in real time and quickly locate the sound source position to meet the real-time needs in the vehicle environment. Its data processing process is simple, and the calculation complexity of sound source positioning is reduced while ensuring the accuracy of sound source positioning. It is easy to implement in the actual system and does not require complex hardware equipment.

[0025] In one embodiment of the present application, the format of the audio data is RAW, the number of channels is 2 or 1, and the sampling frequency is 44100 Hz.

[0026] Specifically, the RAW audio format is the original audio data that has not been compressed and processed, retaining the most complete and authentic sound information without any data loss or quality degradation. Since the RAW audio format does not contain any data or compression information, it can be directly used in various audio processing algorithms without decoding or other preprocessing. This makes the processing process simpler and more efficient. Mono audio data processing is relatively simple and is suitable for single sound source situations, such as the input of a single microphone. The computing resources and storage space required to process mono audio data are relatively small, which is suitable for resource-constrained systems. Stereo audio data contains two channels, which can provide richer spatial sense and sound positioning information. Stereo can provide a better sound quality experience, especially in car audio systems, which can enhance the user's listening experience. 44100Hz is the standard sampling frequency of CD quality, which can capture all frequencies that can be heard by the human ear (20Hz to 20kHz), ensuring high fidelity and clarity of audio data. 44100Hz is a common standard for audio processing and playback devices, with good compatibility. Most audio processing software and hardware support this sampling frequency, which is convenient for system integration and application. The 44100Hz sampling frequency strikes a good balance between sound quality and computing resources.

[0027] In one embodiment of the present application, preprocessing each audio data includes the following steps: Recording the ambient noise of the sound collection device; Performing Fourier transform on the audio data and the ambient noise to obtain an audio data spectrum and an ambient noise spectrum; Calculate and set the coefficient between the audio data spectrum and the ambient noise spectrum; According to the coefficient between the audio data spectrum and the ambient noise spectrum, the ambient noise spectrum is subtracted from the audio data spectrum to generate clean audio data.

[0028] Specifically, select a relatively quiet time period with ambient noise to record, avoiding interference from other sudden sounds. During the selected time period, start the sound collection device and start recording the ambient noise. The recording time should be long enough to obtain a stable noise sample. Save the recorded ambient noise as an independent audio file for subsequent processing.

[0029] The audio data to be processed and the previously recorded ambient noise data are read. Fourier transform is applied to the read audio data and ambient noise data respectively to convert them into frequency domain signals. The Fourier transform algorithm includes fast Fourier transform (FFT). Through Fourier transform, the frequency spectrum of the audio data and the frequency spectrum of the ambient noise are obtained.

[0030] Analyze the audio data spectrum and the ambient noise spectrum to determine their amplitude relationship at each frequency. Based on the analysis results, calculate the coefficient between the audio data spectrum and the ambient noise spectrum. This coefficient reflects the proportion of ambient noise in the audio data. It is conceivable that the coefficient can be adjusted. Set the calculated coefficient to a fixed value for subsequent use.

[0031] According to the set coefficients, the audio data spectrum is subtracted to subtract the ambient noise spectrum from the audio data spectrum. The inverse Fourier transform is applied to the noise-subtracted spectrum to convert the frequency domain signal back to the time domain signal to generate clean audio data. The generated clean audio data is saved as a new audio file for subsequent processing and analysis.

[0032] By adopting the above technical solution, the quality of audio data is significantly improved by subtracting environmental noise, making the sound clearer. Clean audio data reduces noise interference, enhances signal availability, and makes subsequent processing and analysis more accurate and reliable. Through Fourier transform and spectral subtraction, environmental noise can be efficiently eliminated, improving the efficiency of audio processing.

[0033] In one embodiment of the present application, preprocessing each audio data further includes the following steps: A linear convolution operation is performed on the clean audio data to generate a smooth audio signal to reduce noise in the clean audio data.

[0034] Specifically, obtain clean audio data. According to the characteristics of the audio data and the type of noise, select a corresponding convolution kernel (such as a Gaussian kernel, a mean filter kernel, etc.). The size and shape of the convolution kernel will affect the effect of the convolution operation.

[0035] A linear convolution operation is performed on the clean audio data to generate a smooth audio signal. The linear convolution operation in this application is a prior art and will not be described in detail here.

[0036] By adopting the above technical solution, the audio signal can be effectively smoothed and the noise component can be reduced through linear convolution operation, making the audio signal clearer. Compared with other noise suppression methods, linear convolution can reduce noise while retaining the original signal characteristics, and has higher signal fidelity. The linear convolution algorithm is relatively simple and easy to implement, and the operation can be accelerated through methods such as FFT to improve processing efficiency.

[0037] In one embodiment of the present application, preprocessing each audio data further includes the following steps: Set the gain factor; The smooth audio signal is gain-adjusted according to the gain factor to increase the signal strength of the smooth audio signal.

[0038] Specifically, a gain factor G is set according to specific application requirements and the characteristics of the audio signal. The gain factor can be a fixed value or dynamically adjusted according to the real-time characteristics of the audio signal. Common setting methods include: A fixed gain value is preset, for example, G=2.

[0039] The gain value is dynamically adjusted according to the average amplitude or peak value of the audio signal to ensure that the signal is within the appropriate range. The smoothed audio signal is obtained from the previous processing step. The signal has been subjected to linear convolution operation and has less noise components.

[0040] Gain adjustment is performed on each audio signal sample, and the gain-adjusted audio signal is converted back to an analog signal and saved as an audio file. Ensure that the amplitude of the output signal is within the dynamic range of the audio device to avoid signal distortion.

[0041] By adopting the above technical solution, the strength of the audio signal can be significantly improved through gain adjustment, making the audio signal louder and clearer in the playback device. The setting of the gain factor can be adjusted according to specific application requirements to adapt to different types of audio signals and usage scenarios.

[0042] In one embodiment of the present application, preprocessing each audio data further includes the following steps: A noise vector having the same length as the smoothed audio signal is generated, the noise vector is fused with the smoothed audio signal having completed gain, and the signal-to-noise ratio of the fused audio model is optimized.

[0043] Specifically, select an appropriate noise type (such as Gaussian white noise, pink noise, etc.) and determine the characteristics of the noise according to the specific application requirements. Use a random number generation algorithm to generate a noise vector with the same length as the smoothed audio signal. You can use standard library functions or custom noise generation algorithms to generate the required noise vector.

[0044] The noise vector is weightedly fused with the smoothed audio signal after gain, the signal-to-noise ratio (SNR) of the current fused signal is calculated, and the weight coefficients are adjusted or filters (such as adaptive filters) are applied to optimize the SNR. Methods such as linear predictive coding (LPC) can be used to further improve the SNR. Parameters are continuously adjusted through iterative optimization algorithms (such as gradient descent) until a satisfactory SNR level is reached.

[0045] By adopting the above technical solution, the audibility of the audio signal can be enhanced through corresponding noise fusion and signal-to-noise ratio optimization, so that it can still be clearly distinguished in the background noise environment. Fusion of noise can improve the robustness of the audio signal, making it more stable in different environments and reducing the impact of environmental changes. By optimizing the signal-to-noise ratio, noise can be effectively suppressed, and the quality of the audio signal can be improved, making it purer and more natural.

[0046] The present application discloses a vehicle-mounted voice positioning system, comprising: A plurality of sound collecting devices for collecting audio data; and The processing unit processes the collected audio data unit using the vehicle-mounted voice positioning method as described above to obtain the sound source position corresponding to the audio data.

[0047] Specifically, the multiple sound collection devices are microphones, which are distributed in different positions in the car and are used to collect sounds in the car. The processing unit is a central processor or a dedicated audio processing chip, which is connected to all sound collection devices. The audio data collected by all microphones will be transmitted to the processing unit by wire or wireless means. Each microphone collects the surrounding audio signals in real time. Each microphone transmits the collected audio data to the processing unit. The location of the sound source is confirmed by the aforementioned method.

[0048] By adopting the above technical solution, the location of the sound source can be more accurately located through the collaborative work of multiple microphones, which greatly improves the response speed and accuracy of the in-vehicle voice control system. The use of multiple microphones can effectively reduce the impact of noise and interference and improve the accuracy of voice recognition.

[0049] In one embodiment of the present application, the number of the sound collecting devices is four, which are evenly distributed on one side of the car seat.

[0050] The above technical solution uses four microphones, which effectively reduces the number of microphones, simplifies the amount of data, and reduces the difficulty of subsequent calculations. It can be imagined that the four microphones are located on one side of the seat in the car cabin, which is convenient for collecting audio near different seats.

[0051] In one embodiment of the present application, the sound collection device is a passive microphone.

[0052] With the above technical solution, the passive microphone does not require an additional power supply, and its working principle is usually based on capacitance or electromagnetic induction, so the power consumption is extremely low. For the vehicle system, this can reduce power consumption and extend the service life of the vehicle battery. Since the passive microphone has no active electronic components, it has a simple structure, a low failure rate and a long service life. This high reliability is particularly important for the vehicle environment, because the vehicle needs to operate stably in a variety of complex environments. Passive microphones are not sensitive to electromagnetic interference and can work more stably in complex electromagnetic environments than active microphones.

[0053] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. All equivalent structural changes made by using the contents of the present invention specification and drawings under the inventive concept of the present invention, or directly / indirectly applied in other related technical fields are included in the patent protection scope of the present invention.

Claims

1. A vehicle-mounted voice positioning method, characterized in that: The following steps are involved: Define the location of the sound collection device; Acquire the audio data collected by each sound collection device, pre-process each audio data, and generate corresponding data to be located; Each data to be located is normalized, all normalized results are converted into positive values ​​and then summed to generate a total audio energy value, the total audio energy value is divided by the sampling number to generate an average energy value, and the sound collection device with the largest average energy value among the sound collection devices is obtained, which is defined as the sound source position.

2. The vehicle-mounted voice positioning method according to claim 1, characterized in that: The format of the audio data is RAW, the number of channels is 2 or 1, and the sampling frequency is 44100 Hz.

3. The vehicle-mounted voice positioning method according to claim 1, characterized in that: Preprocessing each audio data includes the following steps: Recording the ambient noise of the sound collection device; Performing Fourier transform on the audio data and the ambient noise to obtain an audio data spectrum and an ambient noise spectrum; Calculate and set the coefficient between the audio data spectrum and the ambient noise spectrum; According to the coefficient between the audio data spectrum and the ambient noise spectrum, the ambient noise spectrum is subtracted from the audio data spectrum to generate clean audio data.

4. The vehicle-mounted voice positioning method according to claim 3, characterized in that: Preprocessing each audio data also includes the following steps: A linear convolution operation is performed on the clean audio data to generate a smooth audio signal to reduce noise in the clean audio data.

5. The vehicle-mounted voice positioning method according to claim 4, characterized in that: Preprocessing each audio data also includes the following steps: Set the gain factor; The smooth audio signal is gain-adjusted according to the gain factor to increase the signal strength of the smooth audio signal.

6. The vehicle-mounted voice positioning method according to claim 5, characterized in that: Preprocessing each audio data also includes the following steps: A noise vector having the same length as the smoothed audio signal is generated, the noise vector is fused with the smoothed audio signal having completed gain, and the signal-to-noise ratio of the fused audio model is optimized.

7. A vehicle-mounted voice positioning system, characterized in that: include: A plurality of sound collection devices for collecting audio data; and A processing unit processes the collected audio data unit using the vehicle-mounted voice positioning method as described in claims 1 to 6 to obtain the sound source position corresponding to the audio data.

8. The vehicle-mounted voice positioning system according to claim 7, characterized in that: The number of the sound collecting devices is four and they are evenly distributed on one side of the seat of the car.

9. The vehicle-mounted voice positioning system according to claim 7, characterized in that: The sound collecting device is a passive microphone.

Images